2f:Tfe2,You can explore the dataset by editing the SQL query! The fields should be mostly self-explanatory, but let’s copy the dictionary from [Visualization Publication Data - Data Details](https://sites.google.com/site/vispubdata/data-details):

Data dictionary

* __Conference__: Short title of the conference the paper appeared in. One of {SciVis, InfoVis, VAST, Vis}. * __Year__: the year the paper appeared at the conference (note that this can be different from the year the paper has been published as, for example, most VIS 2018 papers were published in the first 2019 issue of the TVCG journal). * __Title__: the title of the paper. * __DOI__: if we found it, a DOI to the paper. Where a paper did not include a valid DOI (2 cases at this point), a syntactically valid but fake DOI was entered that starts with 10.0000. This column can, thus, be used as a unique identifier for each paper. * __Link__: a link to the paper in the IEEE digital library - based on the DOI. * __First Page__: the first page of the paper in the printed proceedings. This information has not been thoroughly checked for correctness. * __Last Page__: the last page of the paper in the printed proceedings. This information has not been thoroughly checked for correctness. * __PaperType__: one of C (conference paper), J (journal paper), M (miscellaneous). If you would like to filter for scientific research articles, then take all Cs and Ts. Papers with an "M" are not typically considered archival publications. * __Abstract__: the abstract of the paper. * __AuthorNames-Deduped__: co-authors are separated by a semicolon. Authors are ordered by how they appear on the paper. The full author names were taken from the de-duped names the [DBLP](https://dblp.uni-trier.de) provides. * __Authors__: author names as they appeared on the paper. Currently this column is not yet complete. For analysis you should always use column AuthorNames-Deduped. * __AuthorAffiliation__: author affiliation. This information has not been cleaned up and for most entries just shows the affiliation of the first author. This information has not been thoroughly checked for correctness. * __InternalReferences__: a list of internal references to other IEEE VIS papers using DOIs of the papers in this list. This list does not include citations to papers external to this conference series. These internal citations are automatically extracted from a full-text extraction of the citation data using [Grobid](https://github.com/kermitt2/grobid) - therefore errors may have occurred. * __AuthorKeywords__: a list of author keywords freely chosen by the authors - extracted as found on the paper pdf. * __AminerCitationCount__: a number indicating how many times this publication has been cited. The citation count was extracted from the [Aminer](https://www.aminer.org/citation) dataset (DBLP-Citation-network V14). The title of the column indicates the last time this count has been updated. A missing value indicates that this publication was not found in the Aminer dataset. * __CitationCount_CrossRef__: a number indicating how many times this publication has been cited. This citation count was extracted from [CrossRef](https://www.crossref.org). A missing value indicates that the article was not found on CrossRef. * __PubsCited__: a number indicating a reference count = the number of other papers this publication cites. This is also automatically extracted via the CrossRef API. * __Award__: a code indicating any VIS internal award the publication has won. BP = best paper, HM = best paper honorable mention (= like a runner-up), TT = test of time award (=an award for a paper that has proved to be influential to the community over time), BA=best application paper award,BCS=best case study award. The data for BP and HM is reliable from about 2004 for InfoVis and 2009 for Vis. Historic records on awards in early years is very sparse (e.g. comes from people's personal websites or CVs). TT awards are a recent addition to the conference, and therefore, complete. 30:T409,htl.html`${d3 .rollups( d3.sort( papers.filter((d) => d["AminerCitationCount"] > 10), (d) => -d.Year ), (v) => d3.sort(v, (d) => -d["AminerCitationCount"]).slice(0, 5), (d) => d.Year ) .map( ([year, table]) => htl.html`

${year}

${Inputs.table(table, { width: {Title: 250, AuthorNames: 200, Award: 30, PaperType: 30}, format: {Link: d => htl.html`${d}`}, columns: [ // "Conference", // "Year", "Title", "AuthorNames", "AminerCitationCount", "CitationCount_CrossRef", //"PubsCited", "Award", //"DOI", "Link", // "FirstPage", // "LastPage", "PaperType" // "AuthorNames-Deduped", //"AuthorAffiliation", // "InternalReferences", // "AuthorKeywords", // "Abstract" ] })}` )}`31:T542,{ const data = papers.map((d, i) => ({ ...d, y: d.Year + ((i % 20) - 10) / 40, // jiggle cited: +d["AminerCitationCount"] || 0 })); const index = d3.index(data, (d) => d.DOI); const links = data.flatMap( (d) => d.InternalReferences?.split(";").map((ref) => ({ ...d, x1: d.cited, x2: index.get(ref)?.cited, y2: index.get(ref)?.y })) || [] ); return Plot.plot({ width, height: 1200, marks: [ Plot.arrow(links, { x1: "x1", x2: "x2", y1: "y", y2: "y2", bend: true, strokeWidth: 1 / 16 }), Plot.dot(data, { x: "cited", y: "y", title: d => `${d.Title}\n${(d["AuthorNames"]||"").split(";").join(", ")}`, href: "Link", target: "_blank", fill: colorBy === "Conference" ? "Conference" : colorBy === "Paper type" ? "PaperType" : "currentColor" }) ], x: { label: "Citations →", labelAnchor: "left", type: "pow", exponent: 0.3, transform: d => d + 1, tickRotate: 30 }, y: { tickFormat: "", grid: true }, color: { legend: true, tickFormat: d => (colorBy === "Conference" ? d : paperType(d)), ...(colorBy === "Conference" && { scheme: "plasma" }) } }); }32:T427,new_author_dict = { let authors_counts = {} let nodes = [] let edges = [] for (let paper in papers){ // console.log(papers[paper]) if (papers[paper] == null) continue; if (papers[paper]["AuthorNames-Deduped"] == null) continue; if (papers[paper]["CitationCount_CrossRef"] < 20) continue; let authors_of_this_paper = papers[paper]["AuthorNames-Deduped"].split(";") if (authors_of_this_paper.length < 2) continue; for (let author of authors_of_this_paper){ if (authors_counts[author] == undefined) authors_counts[author] = 0 authors_counts[author] += 1 } } let new_author_dict = {} for (let author in authors_counts){ if (authors_counts[author] >= 3) new_author_dict[author] = authors_counts[author]; } console.log(Object.keys(new_author_dict).sort((a, b) => a.toLowerCase().localeCompare(b.toLowerCase())) .filter((value, index, self) => index === self.findIndex((t) => t.toLowerCase() === value.toLowerCase()) )) console.log(Object.keys(new_author_dict)) return new_author_dict; }33:T5a5,{ // create our outer SVG element with a size of 500x100 and select it const svg = d3.create('svg') .attr('width', 500) .attr('height', 2000); let count = 0 for (let i in new_author_dict){ svg.append("text") .text(i + " " + new_author_dict[i]) .attr("x", 200) .attr("y", count * 15) .style("text-anchor", "end") count += 1 } let h_index = 210; for (let paper in papers){ if (papers[paper] == null) continue; if (papers[paper]["AuthorNames"] == null) continue; let authors_of_this_paper = papers[paper]["AuthorNames"].split(";") for (let i = 0; i _Dear Philippe, You might be interested in using the\n> [Visualization Publication Dataset](https://sites.google.com/site/vispubdata/home)\n> created by Petra Isenberg. This is the time of year when researchers get to\n> nominate colleagues for awards, or test-of-time awards are selected._\n\nVisualizing a dataset of research in visualization sounded a bit meta, but\nintriguing… With a few colleagues at Observable, who always love doing Dataviz\npairing sessions, we downloaded the dataset and tried to find how to use it.\n\nThen we realized, we’re dataviz practitioners, not experts of this academic\nfield. But why not set up a notebook which would serve as a starting point for\nanyone who wants to explore the data?\n","pinned":false,"hidden":false,"mode":"md"},{"id":907,"value":"---\n\n## The data\n\nThe data is initially offered as a CSV file (**version 10**, downloaded 20 June, 2022). We could load\nit as an array, but it seemed more versatile to instantiate a SQL database with\nit—which we did with the amazing [DuckDB client](https://observablehq.com/@cmudig/duckdb-client) created by the [Data Interaction\nGroup](https://dig.cmu.edu) at Carnegie Mellon University:\n","pinned":false,"hidden":false,"mode":"md"},{"id":906,"value":"import {DuckDBClient} from \"@cmudig/duckdb\"","pinned":false,"hidden":false,"mode":"ojs"},{"id":905,"value":"db = DuckDBClient.of([[\"papers\", FileAttachment(\"IEEE VIS papers 1990-2021 - Main dataset.csv\")]])","pinned":true,"hidden":false,"mode":"ojs"},{"id":1047,"value":"papers = db","pinned":true,"hidden":false,"mode":"ojs"},{"id":903,"value":"$2f","pinned":false,"hidden":false,"mode":"md"},{"id":902,"value":"Also, if you prefer working with good old JavaScript arrays, here’s the data, coming from the SQL cell above:","pinned":false,"hidden":false,"mode":"md"},{"id":992,"value":"papers","pinned":true,"hidden":false,"mode":"ojs"},{"id":900,"value":"---\n## The venues","pinned":false,"hidden":false,"mode":"md"},{"id":899,"value":"The ${d3.format(\",\")(papers.length)} papers come from 4 conferences organized by the IEEE, which have overlapped in time ([more details](https://sites.google.com/site/vispubdata/data-details)); note that, in 2021, all the conferences have finally merged under the unique name VIS. We will be using [Observable Plot](/plot) to draw all our charts.","pinned":false,"hidden":false,"mode":"md"},{"id":898,"value":"venues = Plot.dot(\n papers,\n Plot.group({ r: \"count\" }, { x: \"Year\", y: \"Conference\" })\n).plot({\n marginLeft: 60,\n x: { tickFormat: \"\", inset: 10 },\n y: { domain: [\"Vis\", \"InfoVis\", \"VAST\", \"SciVis\"] }\n})","pinned":true,"hidden":false,"mode":"ojs"},{"id":897,"value":"xDomain = venues.scale(\"x\").domain","pinned":true,"hidden":false,"mode":"ojs"},{"id":896,"value":"conferences = venues.scale(\"y\").domain","pinned":true,"hidden":false,"mode":"ojs"},{"id":895,"value":"---\n## The authors","pinned":false,"hidden":false,"mode":"md"},{"id":894,"value":"The dataset is a collection of papers, which have multiple authors. The field\n“AuthorNames-Deduped” seems like the cleanest way to retrieve all the authors\nand their papers. This histogram shows the distribution of number of\ncollaborators per paper, depending on the type of paper:\n","pinned":false,"hidden":false,"mode":"md"},{"id":893,"value":"Plot.plot({\n height: 180,\n x: { domain: d3.range(1, 18), label: \"# of coauthors\", line: true },\n y: { nice: true, grid: true },\n color: {\n legend: true,\n tickFormat: paperType\n },\n marks: [\n Plot.barY(\n papers,\n Plot.groupX(\n { y: \"count\" },\n {\n x: (d) => d[\"AuthorNames-Deduped\"].split(\";\").length,\n fill: \"PaperType\",\n order: \"sum\"\n }\n )\n )\n ]\n})","pinned":true,"hidden":false,"mode":"ojs"},{"id":892,"value":"paperType = (type) =>\n ({ C: \"Conference paper\", J: \"Journal paper\", M: \"Misc.\" }[type])","pinned":false,"hidden":false,"mode":"ojs"},{"id":891,"value":"To help create the dataList in the search input, we’ll sort the authors by their total count of citations:","pinned":false,"hidden":false,"mode":"md"},{"id":890,"value":"rankedAuthors = d3.groupSort(\n papers.flatMap((d) => d[\"AuthorNames-Deduped\"].split(\";\").map((author) => ({ author, ...d }))),\n (papers) => -d3.sum(papers, (d) => d[\"AminerCitationCount\"]),\n (d) => d.author\n)","pinned":false,"hidden":false,"mode":"ojs"},{"id":889,"value":"You can restrict the list of papers and the following charts by full-text search:","pinned":false,"hidden":false,"mode":"md"},{"id":888,"value":"viewof byAuthor = Inputs.search(papers, {\n datalist: rankedAuthors.slice(0, 200)\n})","pinned":false,"hidden":false,"mode":"ojs"},{"id":887,"value":"viewof byConf = Inputs.checkbox(conferences, {\n label: \"Conferences\",\n value: conferences\n})","pinned":false,"hidden":false,"mode":"ojs"},{"id":886,"value":"viewof byType = Inputs.checkbox(new Map([\"C\", \"J\", \"M\"].map(k => [ paperType(k), k])), {\n label: \"Paper type\", \n value: [\"C\", \"J\", \"M\"]\n})","pinned":false,"hidden":false,"mode":"ojs"},{"id":885,"value":"Plot.plot({\n marginLeft: 80,\n marks: [\n Plot.dot(\n papers,\n Plot.group(\n {},\n {\n x: \"Year\",\n y: \"Conference\",\n r: 7,\n strokeWidth: 0.25,\n stroke: \"black\",\n fill: \"white\"\n }\n )\n ),\n Plot.dot(\n selectedPapers,\n Plot.group({ fill: \"count\", r: \"count\" }, { x: \"Year\", y: \"Conference\" })\n )\n ],\n x: { tickFormat: \"\", domain: xDomain, insetLeft: 10 },\n y: { domain: conferences, grid: true },\n color: {\n scheme: \"blues\",\n zero: true\n }\n})","pinned":false,"hidden":false,"mode":"ojs"},{"id":884,"value":"viewof selectedPapers = Inputs.table(\n d3\n .sort(byAuthor, (d) => -d[\"AminerCitationCount\"])\n .filter(\n (d) => (byConf.length === 0 || byConf.includes(d.Conference))\n &&(byType.length === 0 || byType.includes(d.PaperType))\n ),\n { format: { Year: String } }\n)","pinned":false,"hidden":false,"mode":"ojs"},{"id":883,"value":"---\n## The keywords","pinned":false,"hidden":false,"mode":"md"},{"id":882,"value":"Authors add an abstract and keywords to their papers, which allow us to do a very gross analysis of the contents (use the search box to restrict to a set of papers, based on the author’s name, or any other search terms):","pinned":false,"hidden":false,"mode":"md"},{"id":881,"value":"viewof byKeywords = Inputs.search(papers)","pinned":false,"hidden":false,"mode":"ojs"},{"id":880,"value":"WordCloud(\n byKeywords\n .map((d) => d.AuthorKeywords)\n .filter((d) => d != null)\n .flatMap((d) => d.toLowerCase().split(/[,;]/g)).slice(0, 1000),\n {\n width,\n height: 500,\n invalidation // a promise to stop the simulation when the cell is re-run\n }\n)","pinned":false,"hidden":false,"mode":"ojs"},{"id":879,"value":"import {WordCloud} from \"@d3/word-cloud\"","pinned":true,"hidden":false,"mode":"ojs"},{"id":878,"value":"---\n## The citations","pinned":false,"hidden":false,"mode":"md"},{"id":877,"value":"It’s difficult to directly measure the influence of a paper. A (sometimes\ncontroversial) proxy metric is the number of citations. In the chart below, we\ndraw a box plot for each year, which tends to show, first, the obvious: recent\npapers are less cited (since it takes time to score citations…), but also that\nthe field has developed a lot since the early years, which tend to be less cited\nthan papers from 2007. _(Note that we had to skew the *x* axis with an exponent\nin order to show more “meat” of the data. This doesn’t fundamentally alter the\nanalysis, as you can see by changing the exponent, but it will select more or\nless points as outliers.)_\n","pinned":false,"hidden":false,"mode":"md"},{"id":876,"value":"viewof pow = Inputs.range([0.1, 1], {value: 0.3, step:0.01, label:\"exponent\"})","pinned":false,"hidden":false,"mode":"ojs"},{"id":875,"value":"Plot.plot({\n height: 612,\n x: { axis: null, inset: 4, zero: true },\n y: { reverse: true, tickSize: 0 },\n marks: [\n Plot.boxX(\n papers.filter((d) => [\"J\", \"C\"].includes(d.PaperType) && d.Year <= 2019),\n {\n x: (d) => Math.pow(d[\"AminerCitationCount\"], pow),\n y: \"Year\",\n title: d => `${d.Title}\\n${(d[\"AuthorNames\"]||\"\").split(\";\").join(\", \")}`\n }\n )\n ]\n})","pinned":true,"hidden":false,"mode":"ojs"},{"id":939,"value":"Can we show the top 5 papers for a given year by number of Viz citations? Well, certainly:","pinned":false,"hidden":false,"mode":"md"},{"id":943,"value":"$30","pinned":true,"hidden":false,"mode":"ojs"},{"id":874,"value":"Finally, we plot all the papers and their citations. Dots are links to the actual paper or reference on IEEExplore. It’s a bit messy, but it makes it very easy to spot the outliers (the most cited paper by far being “[D³ Data Driven Documents](https://ieeexplore.ieee.org/document/6064996),” by Michael Bostock, Vadim Ogievetsky and Jeffrey Heer, in 2011).","pinned":false,"hidden":false,"mode":"md"},{"id":873,"value":"viewof colorBy = Inputs.radio([\"Conference\", \"Paper type\"], {label: \"Color by\"})","pinned":false,"hidden":false,"mode":"ojs"},{"id":872,"value":"$31","pinned":true,"hidden":false,"mode":"ojs"},{"id":871,"value":"Don’t hesitate to fork this notebook to create your own visualizations and\nexplore the dataset—and please send us suggestions or links to your\nexplorations.\n\n\n---\n*Thanks to Catherine Plaisant and Petra Isenberg for suggesting the dataset, to the dataviz team at observable for exploring it with me, and to Dominik Moritz and Max Bo for code suggestions.*","pinned":false,"hidden":false,"mode":"md"},{"id":1056,"value":"$32","pinned":true,"hidden":false,"mode":"ojs"},{"id":1071,"value":"$33","pinned":false,"hidden":false,"mode":"ojs"}],"files":"$W34","stdlib":"1","titling":"auto"}