Text Analysis & NLP - Word Co-occurrence Analysis Demo 1
Special Topics in Business Analytics
Jonathan Krier, MSBA 580, taught by Jimmy Xu
July 07, 2026
library(tidyverse)
library(tidytext)
library(widyr)
library(igraph)
library(ggraph)
library(Matrix)Completed Analysis — Federal Reserve FOMC Communication
For Part 2, I completed the co-occurrence analysis using the Federal
Reserve’s latest FOMC material that I imported into
fomcprojtabl20260617.txt. This section analyzes the
language of the FOMC communication itself. The goal is to identify which
policy terms appear together most often, so I can see whether the text
emphasizes rates, inflation, uncertainty, projections, labor-market
conditions, or other monetary-policy concepts.
Step 1: Load the Raw Text
fomc_lines <- readLines("fomcprojtabl20260617.txt", encoding = "UTF-8")
gsub("^\\s+|\\s+$", "", fomc_lines, useBytes = TRUE)## [1] "{\\rtf1\\ansi\\ansicpg1252\\cocoartf2870"
## [2] "\\cocoatextscaling0\\cocoaplatform0{\\fonttbl\\f0\\fnil\\fcharset0 HelveticaNeue;}"
## [3] "{\\colortbl;\\red255\\green255\\blue255;\\red242\\green242\\blue242;\\red10\\green10\\blue10;}"
## [4] "{\\*\\expandedcolortbl;;\\cssrgb\\c96078\\c96078\\c96078;\\cssrgb\\c3922\\c3922\\c3922;}"
## [5] "\\margl1440\\margr1440\\vieww14780\\viewh22340\\viewkind0"
## [6] "\\deftab720"
## [7] "\\pard\\pardeftab720\\partightenfactor0"
## [8] ""
## [9] "\\f0\\fs24 \\cf2 \\cb3 \\expnd0\\expndtw0\\kerning0"
## [10] "\\outl0\\strokewidth0 \\strokec2 ### For release at 2:00 p.m., EDT, June 17, 2026\\"
## [11] "\\"
## [12] "# Summary of Economic Projections\\"
## [13] "\\"
## [14] "## In conjunction with the Federal Open Market Committee (FOMC) meeting held on\\"
## [15] "\\"
## [16] "## June 16\\'9617, 2026, meeting participants submitted their projections of the most likely\\"
## [17] "\\"
## [18] "## outcomes for real gross domestic product (GDP) growth, the unemployment rate, and\\"
## [19] "\\"
## [20] "## inflation for each year from 2026 to 2028 and over the longer run. Each participant\\'92s\\"
## [21] "\\"
## [22] "## projections were based on information available at the time of the meeting, together\\"
## [23] "\\"
## [24] "## with her or his assessment of appropriate monetary policy\\'97including a path for the\\"
## [25] "\\"
## [26] "## federal funds rate and its longer-run value\\'97and assumptions about other factors likely\\"
## [27] "\\"
## [28] "## to affect economic outcomes. The longer-run projections represent each participant\\'92s\\"
## [29] "\\"
## [30] "## assessment of the value to which each variable would be expected to converge, over\\"
## [31] "\\"
## [32] "## time, under appropriate monetary policy and in the absence of further shocks to\\"
## [33] "\\"
## [34] "## the economy. \\'93Appropriate monetary policy\\'94 is defined as the future path of policy\\"
## [35] "\\"
## [36] "## that each participant deems most likely to foster outcomes for economic activity and\\"
## [37] "\\"
## [38] "inflation that best satisfy his or her individual interpretation of the statutory mandate to promote maximum employment and price stability.\\"
## [39] "\\"
## [40] "## Page of\\"
## [41] "\\"
## [42] "---\\"
## [43] "\\"
## [44] "\\"
## [45] ": Projections of change in real gross domestic product (GDP) and projections for both measures of inflation are percent changes from the fourth\\"
## [46] "quarter of the previous year to the fourth quarter of the year indicated. PCE inflation and core PCE inflation are the percentage rates of change in,\\"
## [47] "respectively, the price index for personal consumption expenditures (PCE) and the price index for PCE excluding food and energy. Projections for the\\"
## [48] "unemployment rate are for the average civilian unemployment rate in the fourth quarter of the year indicated. Each participant\\'92s projections are based on\\"
## [49] "his or her assessment of appropriate monetary policy. Longer-run projections represent each participant\\'92s assessment of the rate to which each variable would\\"
## [50] "be expected to converge under appropriate monetary policy and in the absence of further shocks to the economy. The projections for the federal funds rate\\"
## [51] "are the value of the midpoint of the projected appropriate target range for the federal funds rate or the projected appropriate target level for the federal\\"
## [52] "funds rate at the end of the specified calendar year or over the longer run. The March projections were made in conjunction with the meeting of the Federal\\"
## [53] "Open Market Committee on March 17\\'9618, 2026. Nineteen participants submitted information in conjunction with the March 17\\'9618, 2026, meeting. Eighteen\\"
## [54] "participants submitted information in conjunction with the June 16\\'9617, 2026, meeting; one of these 18 participants did not submit projections for 2028.\\"
## [55] "1. For each period, the median is the middle projection when the projections are arranged from lowest to highest. When the number of projections is\\"
## [56] "3. The range for a variable in a given year includes all participants\\'92 projections, from lowest to highest, for that variable in that year.\\"
## [57] "\\"
## [58] "\\"
## [59] ": Projections of change in real gross domestic product (GDP) and projections for both measures of inflation are percent changes from the fourth\\"
## [60] "quarter of the previous year to the fourth quarter of the year indicated. PCE inflation and core PCE inflation are the percentage rates of change in,\\"
## [61] "respectively, the price index for personal consumption expenditures (PCE) and the price index for PCE excluding food and energy. Projections for the\\"
## [62] "unemployment rate are for the average civilian unemployment rate in the fourth quarter of the year indicated. Each participant\\'92s projections are based on\\"
## [63] "his or her assessment of appropriate monetary policy. Longer-run projections represent each participant\\'92s assessment of the rate to which each variable would\\"
## [64] "be expected to converge under appropriate monetary policy and in the absence of further shocks to the economy. The projections for the federal funds rate\\"
## [65] "are the value of the midpoint of the projected appropriate target range for the federal funds rate or the projected appropriate target level for the federal\\"
## [66] "funds rate at the end of the specified calendar year or over the longer run. The March projections were made in conjunction with the meeting of the Federal\\"
## [67] "Open Market Committee on March 17\\'9618, 2026. Nineteen participants submitted information in conjunction with the March 17\\'9618, 2026, meeting. Eighteen\\"
## [68] "participants submitted information in conjunction with the June 16\\'9617, 2026, meeting; one of these 18 participants did not submit projections for 2028.\\"
## [69] "1. For each period, the median is the middle projection when the projections are arranged from lowest to highest. When the number of projections is\\"
## [70] "even, the median is the average of the two middle projections.\\"
## [71] "2. The central tendency excludes the three highest and three lowest projections for each variable in each year.\\"
## [72] "3. The range for a variable in a given year includes all participants\\'92 projections, from lowest to highest, for that variable in that year.\\"
## [73] "4. Longer-run projections for core PCE inflation are not collected.\\"
## [74] "Page 2 of\\"
## [75] "\\"
## [76] "Table 1. Economic projections of Federal Reserve Board members and Federal Reserve Bank presidents,\\"
## [77] "under their individual assumptions of projected appropriate monetary policy, June 2026\\"
## [78] ": Projections of change in real gross domestic product (GDP) and projections for both measures of inflation are percent changes from the fourth\\"
## [79] "quarter of the previous year to the fourth quarter of the year indicated. PCE inflation and core PCE inflation are the percentage rates of change in,\\"
## [80] "respectively, the price index for personal consumption expenditures (PCE) and the price index for PCE excluding food and energy. Projections for the\\"
## [81] "unemployment rate are for the average civilian unemployment rate in the fourth quarter of the year indicated. Each participant\\'92s projections are based on\\"
## [82] "his or her assessment of appropriate monetary policy. Longer-run projections represent each participant\\'92s assessment of the rate to which each variable would\\"
## [83] "be expected to converge under appropriate monetary policy and in the absence of further shocks to the economy. The projections for the federal funds rate\\"
## [84] "are the value of the midpoint of the projected appropriate target range for the federal funds rate or the projected appropriate target level for the federal\\"
## [85] "funds rate at the end of the specified calendar year or over the longer run. The March projections were made in conjunction with the meeting of the Federal\\"
## [86] "Open Market Committee on March 17\\'9618, 2026. Nineteen participants submitted information in conjunction with the March 17\\'9618, 2026, meeting. Eighteen\\"
## [87] "participants submitted information in conjunction with the June 16\\'9617, 2026, meeting; one of these 18 participants did not submit projections for 2028.\\"
## [88] "1. For each period, the median is the middle projection when the projections are arranged from lowest to highest. When the number of projections is\\"
## [89] "even, the median is the average of the two middle projections.\\"
## [90] "2. The central tendency excludes the three highest and three lowest projections for each variable in each year.\\"
## [91] "3. The range for a variable in a given year includes all participants\\'92 projections, from lowest to highest, for that variable in that year.\\"
## [92] "4. Longer-run projections for core PCE inflation are not collected.\\"
## [93] "\\"
## [94] "\\"
## [95] ": Projections of change in real gross domestic product (GDP) and projections for both measures of inflation are percent changes from the fourth\\"
## [96] "quarter of the previous year to the fourth quarter of the year indicated. PCE inflation and core PCE inflation are the percentage rates of change in,\\"
## [97] "respectively, the price index for personal consumption expenditures (PCE) and the price index for PCE excluding food and energy. Projections for the\\"
## [98] "unemployment rate are for the average civilian unemployment rate in the fourth quarter of the year indicated. Each participant\\'92s projections are based on\\"
## [99] "his or her assessment of appropriate monetary policy. Longer-run projections represent each participant\\'92s assessment of the rate to which each variable would\\"
## [100] "be expected to converge under appropriate monetary policy and in the absence of further shocks to the economy. The projections for the federal funds rate\\"
## [101] "are the value of the midpoint of the projected appropriate target range for the federal funds rate or the projected appropriate target level for the federal\\"
## [102] "funds rate at the end of the specified calendar year or over the longer run. The March projections were made in conjunction with the meeting of the Federal\\"
## [103] "Open Market Committee on March 17\\'9618, 2026. Nineteen participants submitted information in conjunction with the March 17\\'9618, 2026, meeting. Eighteen\\"
## [104] "participants submitted information in conjunction with the June 16\\'9617, 2026, meeting; one of these 18 participants did not submit projections for 2028.\\"
## [105] "1. For each period, the median is the middle projection when the projections are arranged from lowest to highest. When the number of projections is\\"
## [106] "even, the median is the average of the two middle projections.\\"
## [107] "2. The central tendency excludes the three highest and three lowest projections for each variable in each year.\\"
## [108] "3. The range for a variable in a given year includes all participants\\'92 projections, from lowest to highest, for that variable in that year.\\"
## [109] "4. Longer-run projections for core PCE inflation are not collected.\\"
## [110] "\\"
## [111] "under their individual assumptions of projected appropriate monetary policy, June 2026\\"
## [112] "Longer\\"
## [113] "run\\"
## [114] "\\"
## [115] ": Projections of change in real gross domestic product (GDP) and projections for both measures of inflation are percent changes from the fourth\\"
## [116] "quarter of the previous year to the fourth quarter of the year indicated. PCE inflation and core PCE inflation are the percentage rates of change in,\\"
## [117] "respectively, the price index for personal consumption expenditures (PCE) and the price index for PCE excluding food and energy. Projections for the\\"
## [118] "unemployment rate are for the average civilian unemployment rate in the fourth quarter of the year indicated. Each participant\\'92s projections are based on\\"
## [119] "his or her assessment of appropriate monetary policy. Longer-run projections represent each participant\\'92s assessment of the rate to which each variable would\\"
## [120] "be expected to converge under appropriate monetary policy and in the absence of further shocks to the economy. The projections for the federal funds rate\\"
## [121] "are the value of the midpoint of the projected appropriate target range for the federal funds rate or the projected appropriate target level for the federal\\"
## [122] "funds rate at the end of the specified calendar year or over the longer run. The March projections were made in conjunction with the meeting of the Federal\\"
## [123] "Open Market Committee on March 17\\'9618, 2026. Nineteen participants submitted information in conjunction with the March 17\\'9618, 2026, meeting. Eighteen\\"
## [124] "participants submitted information in conjunction with the June 16\\'9617, 2026, meeting; one of these 18 participants did not submit projections for 2028.\\"
## [125] "1. For each period, the median is the middle projection when the projections are arranged from lowest to highest. When the number of projections is\\"
## [126] "even, the median is the average of the two middle projections.\\"
## [127] "2. The central tendency excludes the three highest and three lowest projections for each variable in each year.\\"
## [128] "3. The range for a variable in a given year includes all participants\\'92 projections, from lowest to highest, for that variable in that year.\\"
## [129] "4. Longer-run projections for core PCE inflation are not collected.\\"
## [130] "\\"
## [131] "---\\"
## [132] "\\"
## [133] "Figure 1. Medians, central tendencies, and ranges of economic projections, 2026\\'9628 and over the longer run\\"
## [134] "\\"
## [135] "---\\"
## [136] "\\"
## [137] "Figure 2. FOMC participants\\'92 assessments of appropriate monetary policy: Midpoint of target range\\"
## [138] "or target level for the federal funds rate\\"
## [139] "\\"
## [140] "Note: Each shaded circle indicates the value (rounded to the nearest1/8 percentage point) of an individual\\"
## [141] "participant\\'92s judgment of the midpoint of the appropriate target range for the federal funds rate or the appropriate\\"
## [142] "target level for the federal funds rate at the end of the specified calendar year or over the longer run.\\"
## [143] "\\"
## [144] "^\\{1\\}8\\"
## [145] "\\"
## [146] "---\\"
## [147] "\\"
## [148] "Figure 3.A. Distribution of participants\\'92 projections for the change in real GDP, 2026\\'9628 and over the longer run\\"
## [149] "\\"
## [150] "---\\"
## [151] "\\"
## [152] "Figure 3.B. Distribution of participants\\'92 projections for the unemployment rate, 2026\\'9628 and over the longer run\\"
## [153] "\\"
## [154] "---\\"
## [155] "\\"
## [156] "Figure 3.C. Distribution of participants\\'92 projections for PCE inflation, 2026\\'9628 and over the longer run\\"
## [157] "\\"
## [158] "---\\"
## [159] "\\"
## [160] "Figure 3.D. Distribution of participants\\'92 projections for core PCE inflation, 2026\\'9628\\"
## [161] "\\"
## [162] "Note: Definitions of variables and other explanations are in the notes to table 1.\\"
## [163] "\\"
## [164] "---\\"
## [165] "\\"
## [166] "Figure 3.E. Distribution of participants\\'92 judgments of the midpoint of the appropriate target range for the\\"
## [167] "federal funds rate or the appropriate target level for the federal funds rate, 2026\\'9628 and over the longer run\\"
## [168] "\\"
## [169] "---\\"
## [170] "\\"
## [171] "## For release at 2:00 p.m., EDT, June 17, 2026\\"
## [172] "\\"
## [173] "Figure 4.A. Uncertainty and risks in projections of GDP growth\\"
## [174] "\\"
## [175] "Median projection and confidence interval based on historical forecast errors Percent Change in real GDP\\"
## [176] "\\"
## [177] "6 Actual 5 234 1 0 Median of projections70% confidence interval \\uc0\\u8722 1\\"
## [178] "\\"
## [179] "2021 2022 2023 2024 2025 2026 2027 2028\\"
## [180] "\\"
## [181] "FOMC participants\\'92 assessments of uncertainty and risks around their economic projections\\"
## [182] "\\"
## [183] "Number of Participants Number of Participants Uncertainty about GDP growth Risks to GDP growth June projections 1820 June projections 1820 March projections 16 March projections 16 14 14 12 12 10 10 8 8 6 6 4 4 2 2\\"
## [184] "\\"
## [185] "Lower Broadly Higher Weighted to Broadly Weighted to similar downside balanced upside\\"
## [186] "\\"
## [187] "Note: The blue and red lines in the top panel show actual values and median projected values, respectively, of the\\"
## [188] "\\"
## [189] "percent change in real gross domestic product (GDP) from the fourth quarter of the previous year to the fourth quarter of the year indicated. The confidence interval around the median projected values is assumed to be symmetric and is based on root mean squared errors of various private and government forecasts made over the previous 20 years; more information about these data is available in table 2. Because current conditions may differ from those that prevailed, on average, over the previous 20 years, the width and shape of the confidence interval estimated on the basis of the historical forecast errors may not reflect FOMC participants\\'92 current assessments of the uncertainty and risks around their projections; these current assessments are summarized in the lower panels. Generally speaking, participants who judge the uncertainty about their projections as \\'93broadly similar\\'94 to the average levels of the past 20 years would view the width of the confidence interval shown in the historical fan chart as largely consistent with their assessments of the uncertainty about their projections. Likewise, participants who judge the risks to their projections as \\'93broadly balanced\\'94 would view the confidence interval around their projections as approximately symmetric. For definitions of uncertainty and risks in economic projections, see the box \\'93Forecast Uncertainty.\\'94\\"
## [190] "\\"
## [191] "# Page of\\"
## [192] "\\"
## [193] "---\\"
## [194] "\\"
## [195] "Figure 4.B. Uncertainty and risks in projections of the unemployment rate\\"
## [196] "\\"
## [197] "Median projection and confidence interval based on historical forecast errors\\"
## [198] "\\"
## [199] "FOMC participants\\'92 assessments of uncertainty and risks around their economic projections\\"
## [200] "\\"
## [201] "Note: The blue and red lines in the top panel show actual values and median projected values, respectively, of the\\"
## [202] "average civilian unemployment rate in the fourth quarter of the year indicated. The confidence interval around the\\"
## [203] "median projected values is assumed to be symmetric and is based on root mean squared errors of various private and\\"
## [204] "government forecasts made over the previous 20 years; more information about these data is available in table 2.\\"
## [205] "Because current conditions may differ from those that prevailed, on average, over the previous 20 years, the width\\"
## [206] "and shape of the confidence interval estimated on the basis of the historical forecast errors may not reflect FOMC\\"
## [207] "participants\\'92 current assessments of the uncertainty and risks around their projections; these current assessments are\\"
## [208] "summarized in the lower panels. Generally speaking, participants who judge the uncertainty about their projections as\\"
## [209] "\\'93broadly similar\\'94 to the average levels of the past 20 years would view the width of the confidence interval shown in the\\"
## [210] "historical fan chart as largely consistent with their assessments of the uncertainty about their projections. Likewise,\\"
## [211] "participants who judge the risks to their projections as \\'93broadly balanced\\'94 would view the confidence interval around\\"
## [212] "their projections as approximately symmetric. For definitions of uncertainty and risks in economic projections, see the\\"
## [213] "box \\'93Forecast Uncertainty.\\'94\\"
## [214] "\\"
## [215] "---\\"
## [216] "\\"
## [217] "Figure 4.C. Uncertainty and risks in projections of PCE inflation\\"
## [218] "\\"
## [219] "Median projection and confidence interval based on historical forecast errors\\"
## [220] "\\"
## [221] "Note: The blue and red lines in the top panel show actual values and median projected values, respectively, of\\"
## [222] "the percent change in the price index for personal consumption expenditures (PCE) from the fourth quarter of the\\"
## [223] "previous year to the fourth quarter of the year indicated. The confidence interval around the median projected values\\"
## [224] "is assumed to be symmetric and is based on root mean squared errors of various private and government forecasts\\"
## [225] "made over the previous 20 years; more information about these data is available in table 2. Because current conditions\\"
## [226] "may differ from those that prevailed, on average, over the previous 20 years, the width and shape of the confidence\\"
## [227] "interval estimated on the basis of the historical forecast errors may not reflect FOMC participants\\'92 current assessments\\"
## [228] "of the uncertainty and risks around their projections; these current assessments are summarized in the lower panels.\\"
## [229] "Generally speaking, participants who judge the uncertainty about their projections as \\'93broadly similar\\'94 to the average\\"
## [230] "levels of the past 20 years would view the width of the confidence interval shown in the historical fan chart as largely\\"
## [231] "consistent with their assessments of the uncertainty about their projections. Likewise, participants who judge the risks\\"
## [232] "to their projections as \\'93broadly balanced\\'94 would view the confidence interval around their projections as approximately\\"
## [233] "symmetric. For definitions of uncertainty and risks in economic projections, see the box \\'93Forecast Uncertainty.\\'94\\"
## [234] "\\"
## [235] "---\\"
## [236] "\\"
## [237] "Figure 4.D. Diffusion indexes of participants\\'92 uncertainty assessments\\"
## [238] "\\"
## [239] "---\\"
## [240] "\\"
## [241] "Figure 4.E. Diffusion indexes of participants\\'92 risk weightings\\"
## [242] "\\"
## [243] "---\\"
## [244] "\\"
## [245] "Figure 5. Uncertainty and risks in projections of the federal funds rate\\"
## [246] "\\"
## [247] "assessments of the uncertainty and risks around their projections.\\"
## [248] "* The confidence interval is derived from forecasts of the average level of short-term interest rates in the fourth\\"
## [249] "quarter of the year indicated; more information about these data is available in table 2. The shaded area encompasses\\"
## [250] "less than a 70 percent confidence interval if the confidence interval has been truncated at zero.\\"
## [251] "\\"
## [252] "that may be appropriate to offset the effects of shocks to the economy.\\"
## [253] "The confidence interval is assumed to be symmetric except when it is truncated at zero - the bottom of the lowest\\"
## [254] "target range for the federal funds rate that has been adopted in the past by the Committee. This truncation would\\"
## [255] "not be intended to indicate the likelihood of the use of negative interest rates to provide additional monetary policy\\"
## [256] "accommodation if doing so was judged appropriate. In such situations, the Committee could also employ other tools,\\"
## [257] "including forward guidance and large-scale asset purchases, to provide additional accommodation. Because current\\"
## [258] "conditions may differ from those that prevailed, on average, over the previous 20 years, the width and shape of the\\"
## [259] "confidence interval estimated on the basis of the historical forecast errors may not reflect FOMC participants\\'92 current\\"
## [260] "assessments of the uncertainty and risks around their projections.\\"
## [261] "* The confidence interval is derived from forecasts of the average level of short-term interest rates in the fourth\\"
## [262] "\\"
## [263] "---\\"
## [264] "\\"
## [265] "\\"
## [266] "Note: Error ranges shown are measured as plus or minus the root mean squared\\"
## [267] "error of projections for 2006 through 2025 that were released in the summer by\\"
## [268] "various private and government forecasters. As described in the box \\'93Forecast\\"
## [269] "Uncertainty,\\'94 under certain assumptions, there is about a 70 percent probability\\"
## [270] "that actual outcomes for real GDP, unemployment, consumer prices, and the federal\\"
## [271] "funds rate will be in ranges implied by the average size of projection errors made in\\"
## [272] "the past. For more information, see David Reifschneider and Peter Tulip (2017),\\"
## [273] "\\'93Gauging the Uncertainty of the Economic Outlook Using Historical Forecasting\\"
## [274] "Errors: The Federal Reserve\\'92s Approach,\\'94 Finance and Economics Discussion\\"
## [275] "Series 2017-020 (Washington: Board of Governors of the Federal Reserve System,\\"
## [276] "February), https://dx.doi.org/10.17016/FEDS.2017.020.\\"
## [277] "1. Definitions of variables are in the general note to table 1.\\"
## [278] "\\"
## [279] "1. Definitions of variables are in the general note to table 1.\\"
## [280] "2. Measure is the overall consumer price index, the price measure that has been\\"
## [281] "\\"
## [282] "2. Measure is the overall consumer price index, the price measure that has been\\"
## [283] "most widely used in government and private economic forecasts. Projections are\\"
## [284] "percent changes on a fourth quarter to fourth quarter basis.\\"
## [285] "3. For Federal Reserve staff forecasts, measure is the federal funds rate. For\\"
## [286] "\\"
## [287] "3. For Federal Reserve staff forecasts, measure is the federal funds rate. For\\"
## [288] "other forecasts, measure is the rate on 3-month Treasury bills. Projection errors are\\"
## [289] "calculated using average levels, in percent, in the fourth quarter.\\"
## [290] "\\"
## [291] "---\\"
## [292] "\\"
## [293] "Forecast Uncertainty\\"
## [294] "\\"
## [295] "The economic projections provided by the members of \\"
## [296] "the Board of Governors and the presidents of the Federal \\"
## [297] "Reserve Banks inform discussions of monetary policy among \\"
## [298] "policymakers and can aid public understanding of the basis \\"
## [299] "for policy actions. Considerable uncertainty attends these \\"
## [300] "projections, however. The economic and statistical models \\"
## [301] "and relationships used to help produce economic forecasts \\"
## [302] "are necessarily imperfect descriptions of the real world, and \\"
## [303] "the future path of the economy can be affected by myriad \\"
## [304] "unforeseen developments and events. Thus, in setting the \\"
## [305] "stance of monetary policy, participants consider not only \\"
## [306] "what appears to be the most likely economic outcome as embodied in their projections, but also the range of alternative \\"
## [307] "possibilities, the likelihood of their occurring, and the potential costs to the economy should they occur.\\"
## [308] "Table 2 summarizes the average historical accuracy of a\\"
## [309] "\\"
## [310] "tial costs to the economy should they occur.\\"
## [311] "Table 2 summarizes the average historical accuracy of a \\"
## [312] "range of forecasts, including those reported in past Monetary \\"
## [313] "Policy Reports and those prepared by the Federal Reserve \\"
## [314] "Board\\'92s staff in advance of meetings of the Federal Open \\"
## [315] "Market Committee (FOMC). The projection error ranges \\"
## [316] "shown in the table illustrate the considerable uncertainty associated with economic forecasts. For example, suppose a \\"
## [317] "participant projects that real gross domestic product (GDP) \\"
## [318] "and total consumer prices will rise steadily at annual rates of, \\"
## [319] "respectively, 3 percent and 2 percent. If the uncertainty attending those projections is similar to that experienced in the \\"
## [320] "past and the risks around the projections are broadly balanced, the numbers reported in table 2 would imply a probability of about 70 percent that actual GDP would expand \\"
## [321] "within a range of 1.3 to 4.7 percent in the current year, 1.2 to \\"
## [322] "4.8 percent in the second year, and 0.8 to 5.2 percent in the\\"
## [323] "third year. The corresponding 70 percent confidence intervals for overall inflation would be 1.0 to 3.0 percent in the \\"
## [324] "current year, 0.4 to 3.6 percent in the second year, and 0.6 to \\"
## [325] "3.4 percent in the third year. Figures 4.A through 4.C illustrate these confidence bounds in \\'93fan charts\\'94 that are symmetric and centered on the medians of FOMC participants\\'92 \\"
## [326] "projections for GDP growth, the unemployment rate, and \\"
## [327] "inflation. However, in some instances, the risks around the \\"
## [328] "projections may not be symmetric. In particular, the unemployment rate cannot be negative; furthermore, the risks \\"
## [329] "around a particular projection might be tilted to either the \\"
## [330] "upside or the downside, in which case the corresponding fan \\"
## [331] "chart would be asymmetrically positioned around the median \\"
## [332] "projection.\\"
## [333] "Because current conditions may differ from those that\\"
## [334] "\\"
## [335] "Because current conditions may differ from those that \\"
## [336] "prevailed, on average, over history, participants provide \\"
## [337] "judgments as to whether the uncertainty attached to their \\"
## [338] "projections of each economic variable is greater than, smaller \\"
## [339] "than, or broadly similar to typical levels of forecast uncertainty seen in the past 20 years, as presented in table 2 and \\"
## [340] "reflected in the widths of the confidence intervals shown in\\"
## [341] "the top panels of figures 4.A through 4.C. Participants\\'92 current assessments of the uncertainty surrounding their projec-\\"
## [342] "\\"
## [343] "tions are summarized in the bottom-left panels of those figures. Participants also provide judgments as to whether the \\"
## [344] "risks to their projections are weighted to the upside, are \\"
## [345] "weighted to the downside, or are broadly balanced. That is, \\"
## [346] "while the symmetric historical fan charts shown in the top \\"
## [347] "panels of figures 4.A through 4.C imply that the risks to participants\\'92 projections are balanced, participants may judge that \\"
## [348] "there is a greater risk that a given variable will be above rather \\"
## [349] "than below their projections. These judgments are summarized in the lower-right panels of figures 4.A through 4.C.\\"
## [350] "As with real activity and inflation, the outlook for the\\"
## [351] "\\"
## [352] "rized in the lower-right panels of figures 4.A through 4.C.\\"
## [353] "As with real activity and inflation, the outlook for the \\"
## [354] "future path of the federal funds rate is subject to considerable \\"
## [355] "uncertainty. This uncertainty arises primarily because each \\"
## [356] "participant\\'92s assessment of the appropriate stance of monetary policy depends importantly on the evolution of real activity and inflation over time. If economic conditions evolve \\"
## [357] "in an unexpected manner, then assessments of the appropriate setting of the federal funds rate would change from that \\"
## [358] "point forward. The final line in table 2 shows the error ranges \\"
## [359] "for forecasts of short-term interest rates. They suggest that \\"
## [360] "the historical confidence intervals associated with projections \\"
## [361] "of the federal funds rate are quite wide. It should be noted, \\"
## [362] "however, that these confidence intervals are not strictly consistent with the projections for the federal funds rate, as these\\"
## [363] "projections are not forecasts of the most likely quarterly outcomes but rather are projections of participants\\'92 individual assessments of appropriate monetary policy and are on an endof-year basis. However, the forecast errors should provide a \\"
## [364] "sense of the uncertainty around the future path of the federal \\"
## [365] "funds rate generated by the uncertainty about the macroeconomic variables as well as additional adjustments to monetary \\"
## [366] "policy that would be appropriate to offset the effects of \\"
## [367] "shocks to the economy.\\"
## [368] "If at some point in the future the confidence interval\\"
## [369] "\\"
## [370] "shocks to the economy.\\"
## [371] "If at some point in the future the confidence interval \\"
## [372] "around the federal funds rate were to extend below zero, it \\"
## [373] "would be truncated at zero for purposes of the fan chart \\"
## [374] "shown in figure 5; zero is the bottom of the lowest target \\"
## [375] "range for the federal funds rate that has been adopted by the \\"
## [376] "Committee in the past. This approach to the construction of \\"
## [377] "the federal funds rate fan chart would be merely a convention; \\"
## [378] "it would not have any implications for possible future policy \\"
## [379] "decisions regarding the use of negative interest rates to provide additional monetary policy accommodation if doing so \\"
## [380] "were appropriate. In such situations, the Committee could \\"
## [381] "also employ other tools, including forward guidance and asset \\"
## [382] "purchases, to provide additional accommodation.\\"
## [383] "While figures 4.A through 4.C provide information on}"fomc_text <- tibble(
id = seq_along(fomc_lines),
text = fomc_lines
)
fomc_text## # A tibble: 383 × 2
## id text
## <int> <chr>
## 1 1 "{\\rtf1\\ansi\\ansicpg1252\\cocoartf2870"
## 2 2 "\\cocoatextscaling0\\cocoaplatform0{\\fonttbl\\f0\\fnil\\fcharset0 He…
## 3 3 "{\\colortbl;\\red255\\green255\\blue255;\\red242\\green242\\blue242;\…
## 4 4 "{\\*\\expandedcolortbl;;\\cssrgb\\c96078\\c96078\\c96078;\\cssrgb\\c3…
## 5 5 "\\margl1440\\margr1440\\vieww14780\\viewh22340\\viewkind0"
## 6 6 "\\deftab720"
## 7 7 "\\pard\\pardeftab720\\partightenfactor0"
## 8 8 ""
## 9 9 "\\f0\\fs24 \\cf2 \\cb3 \\expnd0\\expndtw0\\kerning0"
## 10 10 "\\outl0\\strokewidth0 \\strokec2 ### For release at 2:00 p.m., EDT, J…
## # ℹ 373 more rowsEach line in the FOMC file is treated as its own document
(id). That lets the analysis count words as co-occurring
when they appear in the same line of the policy material, which is a
reasonable unit for this formatted FOMC text.
Step 2: Tokenize and Clean
The FOMC file needs more cleaning than the fruit example because it
contains formatting artifacts from the downloaded text, numbers from
tables, and repeated institutional words. I kept meaningful
monetary-policy words such as projections,
rate, inflation, policy, and
uncertainty, while removing generic or technical artifacts
that would distract from the interpretation.
Completed stopword adjustment and reflection
I added custom stopwords that were either institutional labels
repeated throughout the file (fomc, federal,
reserve, june) or formatting artifacts from
the imported text (rtf, ansi,
pard, f0, fs28,
plain, 92s). Removing those words makes the
network more useful because the graph is no longer dominated by
file-format noise or by the name of the institution itself. After this
cleaning, the important terms shift toward actual monetary-policy
content: projections, rate,
participants, uncertainty, funds,
inflation, and policy.
custom_stop_words <- bind_rows(
stop_words,
tibble(
word = c("fomc", "federal", "reserve", "june", "rtf", "ansi", "pard", "f0", "fs28", "plain", "92s"),
lexicon = "custom"
)
)
fomc_words <- fomc_text %>%
mutate(text = str_replace_all(text, "[^[:alnum:][:space:]$]", " ")) %>%
unnest_tokens(word, text, token = "words") %>%
filter(!str_detect(word, "^[0-9]+$")) %>% # drop pure numbers
anti_join(custom_stop_words, by = "word")
fomc_words %>% count(word, sort = TRUE) %>% head(15)## # A tibble: 15 × 2
## word n
## <chr> <int>
## 1 projections 124
## 2 rate 56
## 3 participants 51
## 4 uncertainty 38
## 5 funds 34
## 6 confidence 29
## 7 inflation 29
## 8 pce 28
## 9 policy 27
## 10 fourth 25
## 11 average 24
## 12 interval 24
## 13 monetary 24
## 14 quarter 24
## 15 run 24Step 3: Count Word Pairs
fomc_pairs <- fomc_words %>%
pairwise_count(word, id, sort = TRUE, upper = FALSE)
fomc_pairs %>% head(15)## # A tibble: 15 × 3
## item1 item2 n
## <chr> <chr> <dbl>
## 1 rate funds 33
## 2 projections participants 26
## 3 projections rate 26
## 4 monetary policy 22
## 5 projections risks 20
## 6 projections run 19
## 7 projections uncertainty 19
## 8 confidence interval 19
## 9 conjunction meeting 16
## 10 fourth quarter 16
## 11 projections variable 15
## 12 uncertainty risks 15
## 13 economic projections 14
## 14 rate run 14
## 15 projections lowest 14Step 4: Build the Co-occurrence Matrix
This repeats the fruit-example logic on the FOMC text. Because the policy vocabulary is much larger than the toy fruit dataset, I restrict the matrix to the top 15 frequent words so the output stays readable and the strongest policy themes are easier to compare.
top_words <- fomc_words %>%
count(word, sort = TRUE) %>%
slice_max(n, n = 15) %>%
pull(word)
fomc_matrix <- fomc_pairs %>%
filter(item1 %in% top_words, item2 %in% top_words) %>%
bind_rows(fomc_pairs %>% rename(item1 = item2, item2 = item1) %>%
filter(item1 %in% top_words, item2 %in% top_words)) %>%
cast_sparse(item1, item2, n) %>%
as.matrix()
fomc_matrix## funds participants rate policy run uncertainty interval quarter
## rate 33 1 0 10 14 3 1 6
## projections 12 26 26 11 19 19 3 6
## monetary 6 3 11 22 5 1 0 0
## confidence 1 4 2 0 0 1 19 3
## fourth 0 1 6 0 0 1 5 16
## inflation 0 2 0 1 6 1 0 5
## participants 0 0 1 3 3 10 4 1
## run 8 3 14 5 0 0 0 0
## quarter 0 1 6 0 0 1 3 0
## average 1 3 7 0 0 2 5 8
## policy 5 3 10 0 5 1 0 0
## funds 0 0 33 5 8 2 0 0
## pce 0 2 0 0 5 1 0 6
## uncertainty 2 10 3 1 0 0 1 1
## interval 0 4 1 0 0 1 0 3
## inflation pce monetary fourth average confidence projections
## rate 0 0 11 6 7 2 26
## projections 12 12 11 11 11 5 0
## monetary 1 0 0 0 0 0 11
## confidence 1 0 0 5 6 0 5
## fourth 10 6 0 0 10 5 11
## inflation 0 12 1 10 0 1 12
## participants 2 2 3 1 3 4 26
## run 6 5 5 0 0 0 19
## quarter 5 6 0 16 8 3 6
## average 0 0 0 10 0 6 11
## policy 1 0 22 0 0 0 11
## funds 0 0 6 0 1 1 12
## pce 12 0 0 6 0 0 12
## uncertainty 1 1 1 1 2 1 19
## interval 0 0 0 5 5 19 3Scanning a row of this matrix gives a direct view of the FOMC
communication structure. For example, rate appears strongly
with funds, while monetary appears with
policy, and uncertainty appears with
risks. The network in the next step is a visual version of
those same counts.
Step 5: Visualize the Strongest Co-occurrences
To keep the network readable, I keep only pairs that occurred together more than three times.
set.seed(580)
fomc_pairs %>%
filter(n > 3) %>%
graph_from_data_frame() %>%
ggraph(layout = "fr") +
geom_edge_link(aes(edge_alpha = n, edge_width = n), color = "firebrick") +
geom_node_point(size = 4, color = "steelblue") +
geom_node_text(aes(label = name), repel = TRUE, size = 3.5) +
theme_void() +
labs(
title = "Word Co-occurrence Network: Federal Reserve FOMC Communication",
subtitle = "Edge thickness = number of sentences containing both words"
)
Completed responses
Question 1. Which additional words did you add to the stopword list manually? Why does it make sense to remove those words as stopwords? Did the final network graph change?
I added fomc, federal,
reserve, june, and several formatting
artifacts such as rtf, ansi,
pard, f0, fs28,
plain, and 92s. These words either identify
the source document itself or come from file formatting rather than
policy meaning. Removing them improves the graph because the largest
nodes and edges are now more about the substance of the FOMC material.
The final graph becomes easier to interpret: instead of technical noise,
it highlights terms like projections, rate,
funds, inflation, policy,
uncertainty, and risks.
Question 2. Change n in
filter(n > 1) in Step 5 to a different value. What is
your final value of n? Will it give you a better result?
I would use filter(n > 3) for the final graph. A
threshold of n > 1 keeps too many weak connections and
makes the network crowded. Raising the cutoff to n > 3
keeps repeated relationships while removing one-off or low-frequency
pairs. For this FOMC text, that produces a cleaner business-style
visualization because the remaining edges emphasize recurring themes
such as rate/funds, monetary/policy,
confidence/interval, and
uncertainty/risks.
Additional discussion responses
Question 3. Which co-occurrence network graph style do you like better? Why?
I prefer the cleaner graph with fewer edges because it communicates the pattern faster. A dense network may be technically complete, but it is harder to explain in a business report. For this FOMC analysis, the best graph is the one that makes the repeated policy relationships visible without overwhelming the reader.
Question 4. Which word pairs surprised you?
The strongest and most useful pairs were policy-specific rather than
emotional. rate/funds was expected because the federal
funds rate is central to FOMC communication. I also found
uncertainty/risks and confidence/interval
useful because they show that the document is not only about a rate
decision; it is also about how uncertain participants are around
projections. That matters because the Fed communicates both policy
direction and the range of possible outcomes.
Question 5. Should we treat FOMC,
Federal Reserve, and related labels as the same token or
remove them?
For this analysis, I removed them as custom stopwords instead of treating them as separate meaningful tokens. Because every line comes from the same Federal Reserve/FOMC source, those labels do not help distinguish themes inside the document. Removing them lets the analysis focus on what the FOMC is discussing: projections, rates, inflation, uncertainty, and policy risks.
Question 6. What is the risk of drawing conclusions from a small co-occurrence sample? What sample size would be better?
The main risk is over-interpreting accidental word pairings. A small or highly formatted document can make some pairs look important simply because the same table structure repeats. For public comments or reviews, I would prefer hundreds or thousands of observations before making strong claims. For this FOMC document, I would frame the findings as descriptive: the network summarizes this specific FOMC material, not all Federal Reserve communication or market reaction to it.
Question 7. How could this be applied to comments or reviews around a significant event?
Yes. A useful extension would be to compare comments or news headlines before and after an FOMC announcement. I would identify the event by its official release time, such as the June 17, 2026 FOMC statement and press conference, then collect text from a window before and after the announcement. The event is significant because FOMC communication can change expectations about interest rates, inflation, borrowing costs, and market risk. Comparing co-occurrence networks before and after the event could show whether public discussion shifted from broad uncertainty toward specific concerns like inflation, rate cuts, labor markets, or financial conditions.
Wrap-Up: From Toy Example to Business Insight
The fruit example builds intuition, and the completed FOMC analysis shows why the same workflow is useful on real business text. After cleaning the imported FOMC material, the co-occurrence network highlights recurring monetary-policy concepts: projections, rates, funds, inflation, uncertainty, risks, and policy. The main takeaway is that word co-occurrence helps turn an unstructured policy document into a map of repeated themes that can be interpreted for business or market analysis.
References
Boost Brand Credibility with Co Occurrence SEO. https://www.linkedin.com/posts/umer-abid-78045131a_co-occurrence-seo-is-a-powerful-concept-that-activity-7421759779892809728-MGh_/
https://seopressor.com/blog/why-co-citation-and-co-occurrence-are-such-big-deal/
Kong, J., Scott, A., & Goerg, G. M. (2016). Improving topic clustering on search queries with word co-occurrence and bipartite graph coclustering. https://research.google/pubs/improving-topic-clustering-on-search-queries-with-word-co-occurrence-and-bipartite-graph-co-clustering/
Colladon, A. F. (2018). The semantic brand score. Journal of Business Research, 88, 150-160.
Silge, J., & Robinson, D. (2017). Text Mining with R: A Tidy Approach. O’Reilly Media. https://www.tidytextmining.com/
Robinson, D. (2021). widyr: Widen, process, then re-tidy data [R package documentation]. https://CRAN.R-project.org/package=widyr
Csardi, G., & Nepusz, T. (2006). The igraph software package for complex network research. InterJournal, Complex Systems, 1695. https://igraph.org
Pedersen, T. L. (2024). ggraph: An implementation of grammar of graphics for graphs and networks [R package documentation]. https://CRAN.R-project.org/package=ggraph