2a:null 2b:["$","title",null,{"children":"In Search of Marcel Proust | Guillaume Lesaine | Observable"}] 30:T452,md`**À la Recherche du temps perdu**, written between 1906 and 1922 by **Marcel Proust** is a masterpiece of the French Litterature. It is also a behemoth since it contains : **1 208 000** words, **40 000** sentences, **151 000** punctuation marks, and **7** volumes. This document is thus a great subject for text analysis. In this study we want to better understand the construction of la Recherche. We will follow these steps: * [\`Macro Analysis\`](#macroAnalysis) * [\`Ngrams Analysis\`](#ngramsAnalysis) * [\`Thematic Analysis\`](#thematicAnalysis) Throughout the study we will alternatively focus on the full text and the volumes. For future reference the volumes are numbered from 0 to 6: **0** : **Du côté de chez Swann** (Swann's Way), **1** : **À l’ombre des jeunes filles en fleurs** (In the Shadow of Young Girls in Flower), **2** : **Le Côté de Guermantes** (The Guermantes Way), **3** : **Sodome et Gomorrhe** (Sodom and Gomorrah), **4** : **La Prisonnière** (The Prisoner), **5** : **Albertine disparue** (The Fugitive), **6** : **Le Temps retrouvé** (Time Regained). `31:Ted9,{ const context = { 'height':500, 'width': 1200, 'margin': {'left': 40, 'right': 20, 'top': 40, 'bottom': 40}, 'cLeft': {'h': 1, 'w': 0.80, 'm': {'l': 0.5, 'r': 0.5, 't': 0.5, 'b': 0.5}}, 'cRight': {'h': 1, 'w': 0.20, 'm': {'l': 0.8, 'r': 0.5, 't': 0.5, 'b': 0.5}}, } const { height, width, margin, cLeft, cRight } = context const svg = d3.create("svg").attr("viewBox", [0, 0, width, height]); const volumes_words_ratio = l_volumes.map(function(d){ return d3.sum(sentences_raw.map(function(e){ return e[d] })) }) const max_sentence = 250 const sentences = sentences_raw.map(function(data){ let obj = new Object() for (const key of l_volumes){ obj[key] = data[key]/volumes_words_ratio[key] } obj["variable"] = data["variable"] console.log(obj['variable']) return obj }).filter(d => d.variable <= max_sentence) const l_sentences = Array.from(new Set(sentences.map(function(d){ return d["variable"] }))) const max_words = d3.max(sentences.map(function(data){ return d3.max(l_volumes.map(function(key){ return data[key] })) })) const xScale = d3.scaleLinear() .domain([1, max_sentence]) .range([margin.left*(1 + cLeft.m.l), width*cLeft.w - margin.right*cLeft.m.r])// const yScale = d3.scaleBand() .domain(l_volumes) .range([height*cLeft.h - margin.bottom*(1 + cLeft.m.b), margin.top*(1 + cLeft.m.t)]) .padding(0.2) const colorScale = d3.scaleLinear() .domain([0, max_words]) .range([0, 1]) let xAxis = g => g .attr("transform", `translate(0,${height*cLeft.h - margin.bottom*(1 + cLeft.m.b)})`) .call(d3.axisBottom(xScale)); let yAxis = g => g .attr("transform", `translate(${margin.left*(1 + cLeft.m.l)},0)`) .call(d3.axisLeft(yScale)); function color(x){ if (x === 0){ return "white" } else { return d3.interpolateReds(x) } } let chart = svg.append('g') .attr('id', 'chart') chart.selectAll("g") .data(sentences) .enter() .append("g") .selectAll("rect") .data(d => l_volumes.map(function(e){ return {"value":d[e],"variable":d["variable"]} })) .enter() .append("rect") .attr("x", d => xScale(d["variable"])) .attr("y", (d, i) => yScale(i)) .attr("width", (xScale.range()[1] - xScale.range()[0])/max_sentence) .attr("height", yScale.bandwidth() ) .attr("fill", d => color(colorScale(d["value"]))) chart.append("g") .call(xAxis) .append("text") .attr("class", "axis-label") .attr("transform", `translate( ${(width*cLeft.w + margin.left*(1 + cLeft.m.l) - margin.right*(1 + cLeft.m.r)) / 2}, ${2 * margin.bottom*(1 + cLeft.m.b) / 3} )`) .attr("font-size", 14) .text("Words in the sentence") chart.append("g") .call(yAxis) .append("text") .attr("class", "axis-label") .attr("transform", 'rotate(-90)') .attr("x", - (height*cLeft.h / 2)) .attr("y", - (2 * margin.left*(1 + cLeft.m.l) / 3)) .attr("font-size", 14) .text("Volume of la Recherche") chart.append("g") .append("text") .attr("class","chart-title") .attr("transform", `translate(${margin.left*(1 + cLeft.m.l)}, ${margin.top*(1 + cLeft.m.t) / 3})`) .text("Frequence of sentences depending on words number for the volumes of la Recherche") chart.append("g") .append("text") .attr("class","chart-subtitle") .attr("transform", `translate(${margin.left*(1 + cLeft.m.l)}, ${2 * margin.top*(1 + cLeft.m.t) / 3})`) .text("Sentences over 250 words are not represented for convenience.") render_legend_color_range(svg, context, color, [max_words, 0]) return svg.node() }32:T950,md`If you click the above button, you can see groups between main characters. We can thus isolate six sequences: ***Youth*** - **Swann** meets **Odette**, falls in love with her while listening to the composer **Vinteuil**. As his love develops, he is jealous of **Forcheville**, whom he think is having an affair with Odette. **Gilberte** is the daughter of **Swann** and **Odette**, with whom the narrator plays during his childhood. - Childhood and youth of the narrator with the maid **Françoise**, episodes with an actress he admires **Berma** and his favorite writer **Bergotte**. **Norpois** is a diplomat, dear friend of the father of the narrator. ***Nobility*** - The narrator goes to Balbec for holidays in Normandy with his grand-mother. There he meets **Robert de Saint-Loup** and **Mme de Villeparisis**, both closely linked to the noble family of Guermantes. **Bloch** a "snob" friend of the narrator visits him in Balbec at one point. - The third volume is entitled "Le Coté de Guermantes" (The Guermantes Way), it happens in Paris and is mainly about the fascination the narrator vows to the family of **Guermantes**, and especially the Duchess of Guermantes. ***Girls in flower*** - Moment of la Recherche where the narrator meets a group of young girls in Balbec. Among them are **Albertine** and **Andrée**. He cannot figure out which one he loves best. They all meet at **Elstir**'s place who is a painter. ***Mundanity*** - The second half of the book is marked by the story of the Baron de **Charlus**, a homosexual nobleman from the family of Guermantes. He is desperately in love with a young violonist called Charles **Morel**. They both spend nights with the **Verdurins** and a gathering of middle to upper class snobs like **Cottard**, **Brichot** and **Cambremer**. ***Love*** - The narrator finally falls in love with **Albertine**, one of the girls he met before. She comes to his place in Paris. She dies before the beginning of the last volume. ***Guermantes' Reception*** - La Recherche ends with a long reception held at the **Guermantes**'. There, a lot of characters of the book meet. The narrator meets again **Gilberte**, **Swann**'s daughter. **Bloch**, the "snob" from the second volume became an admired writer after all those years. This is, to me, an accurate summary of the thousands of pages of la Recherche. `33:T108e,function sentences_chart() { const margin = { top: 50, right: 10, bottom: 70, left: 70 }; const w = 600; const h = 600; const sentences = sentences_raw.map(function(data) { let obj = new Object(); obj["variable"] = data["variable"]; obj["count"] = d3.sum(l_volumes.map(d => data[d])); return obj; }); let sum_words = d3.sum(sentences.map(d => d.variable * d.count)); let vector = [0.25, 0.5, 0.75, 1]; let parts = vector.map(function(p) { let i = 0; let sum = sentences[i]["variable"] * sentences[i]["count"]; while (sum <= p * sum_words) { sum = sum + sentences[i]["variable"] * sentences[i]["count"]; i += 1; } return sentences[i]["variable"]; }); const svg_scatter_plot = d3.create("svg").attr("viewBox", [0, 0, w, h]); let xScale = d3 .scaleLog() .domain([0.9, d3.max(sentences.map(data => data["variable"]))]) .range([margin.left, w - margin.right]); let max_count = d3.max(sentences.map(data => data["count"])); let yScale = d3 .scaleLinear() .domain([-10, max_count]) .range([h - margin.bottom, margin.top]); let colorScale = d3.scaleOrdinal(d3.schemeCategory10); let xAxis = g => g .attr("transform", `translate(0,${h - margin.bottom})`) .call(d3.axisBottom(xScale)); let yAxis = g => g .attr("transform", `translate(${margin.left},0)`) .call(d3.axisLeft(yScale).ticks(h / 80)); svg_scatter_plot .selectAll("line") .data(parts.slice(0, -1)) .enter() .append("line") .attr("x1", d => xScale(d)) .attr("y1", h - margin.bottom) .attr("x2", d => xScale(d)) .attr("y2", (d, i) => yScale(max_count) + 70 * i) .style("stroke-width", 1.5) .style("stroke-dasharray", "3, 3") .style("stroke", (d, i) => colorScale(i)); svg_scatter_plot .append("g") .selectAll("text") .data(parts.slice(0, -1)) .enter() .append("text") .attr("x", d => xScale(d) + 10) .attr("y", (d, i) => yScale(max_count) + 70 * i + 10) .text((d, i) => "< " + (d + 1) + " words") .attr("font-size", 13) .attr("font-weight", 600); svg_scatter_plot .append("g") .selectAll("text") .data(parts.slice(0, -1)) .enter() .append("text") .attr("x", d => xScale(d) + 10) .attr("y", (d, i) => yScale(max_count) + 70 * i + 30) .text((d, i) => vector[i] * 100 + "% of words") .attr("font-size", 13); svg_scatter_plot .append("g") .selectAll("text") .data(parts.slice(0, -1)) .enter() .append("text") .attr("x", d => xScale(d) + 10) .attr("y", (d, i) => yScale(max_count) + 70 * i + 50) .text( (d, i) => Math.round( (d3.sum( sentences .filter(data => data["variable"] <= d) .map(data => data["count"]) ) / number_sentences) * 100 ) + "% of sentences" ) .attr("font-size", 13); svg_scatter_plot .append('g') .selectAll("circle") .data(sentences) .enter() .append("circle") .attr("cx", d => xScale(d["variable"])) .attr("cy", d => yScale(d["count"])) .attr("r", 1.5) .attr("fill", function(d) { let i = 0; while (d["variable"] > parts[i]) { i += 1; } return colorScale(i); }); svg_scatter_plot .append("g") .call(xAxis) .append("text") .attr("class", "axis-label") .attr("transform", `translate(${(w + margin.left - margin.right) / 2}, ${2 * margin.bottom / 3})`) .attr("font-size", 14) .style("text-anchor", "middle") .text("Words in the sentence") svg_scatter_plot .append("g") .call(yAxis) .append("text") .attr("class", "axis-label") .attr("transform", "rotate(-90)") .attr("x", - (h / 2)) .attr("y", - (2 * margin.left / 3)) .attr("font-size", 14) .style("text-anchor", "middle") .text("Number of sentences in la Recherche") svg_scatter_plot .append("g") .append("text") .attr("class", "chart-title") .attr("transform", `translate(${margin.left}, ${margin.top / 2})`) .text(`Number of sentences depending on number of words`); return svg_scatter_plot.node(); }34:T84e,function bar_chart(){ const margin = { top: 50, right: 10, bottom: 70, left: 70 }; const w = 600; const h = 600; const punctuation = punctuation_raw.map(function(data){ let obj = new Object() obj["expression"] = data["expression"] obj["count"] = data["count"]/number_sentences return obj }) const svg_bar_chart = d3.create("svg").attr("viewBox", [0, 0, w, h]); let xScale = d3.scaleBand() .domain(punctuation_raw.map(data => data["expression"])) .range([margin.left, w - margin.right]) .padding(0.3) let yScale = d3.scaleLinear() .domain([0, d3.max(punctuation.map(data => data["count"]))]) .range([h - margin.bottom, margin.top]) let xAxis = g => g .attr("transform", `translate(0,${h - margin.bottom})`) .call(d3.axisBottom(xScale)); let yAxis = g => g .attr("transform", `translate(${margin.left},0)`) .call(d3.axisLeft(yScale).ticks(h / 80)); svg_bar_chart.selectAll("rect") .data(punctuation) .enter() .append("rect") .attr("x", d => xScale(d["expression"])) .attr("y", d => yScale(d["count"])) .attr("width", xScale.bandwidth()) .attr("height", d => yScale(0) - yScale(d["count"])) .attr("fill", "steelblue") svg_bar_chart .append("g") .call(xAxis) .append("text") .attr("class", "axis-label") .attr("transform", `translate(${(w + margin.left - margin.right) / 2}, ${2 * margin.bottom / 3})`) .attr("font-size", 14) .style("text-anchor", "middle") .text("Punctuation mark") svg_bar_chart .append("g") .call(yAxis) .append("text") .attr("class", "axis-label") .attr("transform", "rotate(-90)") .attr("x", - (h / 2)) .attr("y", - (2 * margin.left / 3)) .attr("font-size", 14) .style("text-anchor", "middle") .text("Ratio per sentence") svg_bar_chart .append("g") .append("text") .attr("class", "chart-title") .attr("transform", `translate(${margin.left}, ${margin.top / 2})`) .text(`Ratio of punctuation marks per sentence`); return svg_bar_chart.node() }35:T1030,function violin_chart(obj, title){ const margin = { top: 70, bottom: 70, left: 70, right: 50 } const width = 950 const height = 500 const svg_violin_chart = d3.create("svg").attr("viewBox", [0, 0, width, height]) let xScale = d3.scaleBand() .domain(Object.keys(obj)) .range([margin.left, width - margin.right]) .padding(0.05) let yScale = d3.scaleLinear() .domain([0, 40000]) .range([height - margin.bottom, margin.top]) let xAxis = d3.axisBottom() .scale(xScale) let yAxis = d3.axisLeft() .scale(yScale) let colorScale = d3.scaleOrdinal(d3.schemeCategory10) let histogram = d3.histogram() .domain(yScale.domain()) .thresholds(yScale.ticks(400)) .value(d => d) let obj_bins = Object.keys(obj).map(function(label){ let values = Object.keys(obj[label]).map(function(volume){ return histogram(obj[label][volume]) }) return {"key": label, "values": values} }) let maxBin = d3.max( obj_bins.map(function(label_obj){ return d3.max( label_obj["values"] .map(d => d3.max(d.map(d => d.length))) ) }) ) let widthScale = d3.scaleLinear() .range([0, xScale.bandwidth()]) .domain([-maxBin,maxBin]) svg_violin_chart.selectAll("g") .data(obj_bins) .enter() .append("g") .attr("transform", function(d){ return("translate(" + xScale(d["key"]) +" ,0)") } ) .selectAll("g") .data(d => d["values"]) .enter() .append("g") .append("path") .style("fill", (d,i) => colorScale(i)) .attr("d", d3.area() .x0(function(d){ return(widthScale(-d.length)) } ) .x1(function(d){ return(widthScale(d.length)) } ) .y(function(d){ return(yScale(d.x0)) } ) .curve(d3.curveCatmullRom) ) svg_violin_chart.selectAll("line") .data(sentences_break) .enter() .append("line") .attr("x1", margin.left) .attr("y1", d => yScale(d)) .attr("x2", width - margin.right) .attr("y2", d => yScale(d)) .style("stroke-width", 1.5) .style("stroke-dasharray", ("3, 3")) .style("stroke", (d, i) => colorScale(i)) svg_violin_chart.selectAll("text") .data(sentences_break) .enter() .append("text") .attr("fill", (d, i) => colorScale(i)) .attr("x", width - margin.right + 10) .attr("y", d => yScale(d)) .attr("font-weight", 700) .attr("font-size", 13) .text((d, i) => i) svg_violin_chart.append("g") .attr("class", "axis") .attr("transform", "translate(" + (0) + ", " + (height - margin.bottom) + ")") .call(xAxis) .selectAll("text") .style("text-anchor", "end") .attr("dx", "-0.9em") .attr("dy", ".15em") .attr("transform", "rotate(-30)" ); svg_violin_chart.append("text") .attr("fill", "black") .attr("transform","translate(" + (width/2) + ", " + (height - margin.bottom/4) + ")") .attr("font-weight", 500) .attr("font-size", 13) .attr("font-family","sans-serif") .text("Expressions") .style("text-anchor", "middle"); svg_violin_chart.append("g") .attr("transform", "translate(" + (margin.left) + "," + (0) + ")") .call(yAxis) .append("text") .attr("class", "axis-label") .attr("fill", "black") .attr("transform", "rotate(-90)") .attr("x", -height/2) .attr("y", -margin.left + 10) .attr("font-weight", 500) .attr("font-size", 13) .text("Sentences") .style("text-anchor", "middle") svg_violin_chart.append("text") .attr("class", "axis-label") .attr("fill", "black") .attr("transform", "rotate(90)") .attr("x", height/2) .attr("y", -(width) + 20) .attr("font-weight", 500) .attr("font-size", 13) .attr("font-family","sans-serif") .text("Volumes") .style("text-anchor", "middle") svg_violin_chart.append("g") .append("text") .attr("class","chart-title") .attr("transform", `translate(${margin.left}, ${margin.top / 2})`) .text(title) return {"xScale": xScale, "yScale": yScale, "svg": svg_violin_chart} } 36:T435,function render_legend_color_range(svg, { height, width, margin, cLeft, cRight }, colorScale, values){ let part_number = 300 let data = values const legend = svg.append("g") .attr('id', 'legend') .attr("transform",`translate(${width*cLeft.w}, ${0})`) let parts = d3.range(0, part_number, 1) const legendScale = d3.scaleLinear() .domain([1, part_number]) .range([height*cLeft.h - margin.bottom*(1 + cRight.m.b), margin.top*(1 + cRight.m.t)]) legend.selectAll("rect") .data(parts) .enter() .append("rect") .attr("x", margin.left*cRight.m.l) .attr("y", d => legendScale(d)) .attr("width", 30) .attr("height", 3) .attr("fill", d => colorScale(d/part_number)) legend.selectAll("text") .data(data) .enter() .append("text") .attr('class', 'legend-text') .attr("x", margin.left*cRight.m.l + 40) .attr("y", (d,i) => margin.top*(1 + cRight.m.t) + i*(height*cLeft.h - margin.bottom*(1 + cRight.m.b) - margin.top*(1 + cRight.m.t))) .text(d => Math.round(d * 1000) / 1000) return legend }37:Tf53,function matrix_chart(theme_object, title){ const context = { 'height': 500, 'width': 1000, 'margin': {'left': 40, 'right': 20, 'top': 40, 'bottom': 40}, 'cLeft': {'h': 1, 'w': 0.80, 'm': {'l': 0.5, 'r': 0.5, 't': 0.5, 'b': 0.5}}, 'cRight': {'h': 1, 'w': 0.20, 'm': {'l': 0.8, 'r': 0.5, 't': 0.5, 'b': 0.5}}, } const { height, width, margin, cLeft, cRight } = context const svg_matrix_chart = d3.create("svg").attr("viewBox", [0, 0, width, height]); const theme_labels = Array.from(new Set(theme_object.map(function(d){ return d["expression"] }))) let volumes = l_volumes let max_number = d3.max(theme_object,function(d){ return d3.max(volumes.map(e => d[e])) }) let xScale = d3.scalePoint() .domain(theme_labels) .range([margin.left*(1 + cLeft.m.l), width*cLeft.w - margin.right*cLeft.m.r]) .padding(0.7) let yScale = d3.scalePoint() .domain(volumes) .range([height*cLeft.h - margin.bottom*(1 + cLeft.m.b), margin.top*(1 + cLeft.m.t)]) .padding(0.6) let colorScale = d3.scaleLinear() .domain([0, max_number]) .range([0, 1]) function color(x){ if (x === 0){ return "white" } else { return d3.interpolatePuBu(x) } } let xAxis = g => g .attr("transform", `translate(0,${height*cLeft.h - margin.bottom*(1 + cLeft.m.b)})`) .call(d3.axisBottom(xScale)); let yAxis = g => g .attr("transform", `translate(${margin.left*(1 + cLeft.m.l)},0)`) .call(d3.axisLeft(yScale)); let width_rect = xScale(theme_object[1]["expression"]) - xScale(theme_object[0]["expression"]) let height_rect = yScale(volumes[0]) - yScale(volumes[1]) let chart = svg_matrix_chart.append('g') .attr('id', 'chart') let rectangles = chart.append('g') .selectAll("g") .data(theme_object) .enter() .append("g") .selectAll("rect") .data(d => volumes.map(function(e){ return {"values":d[e],"expression":d["expression"]} })) .enter() .append("rect") .attr("x", d => xScale(d["expression"]) - width_rect/2) .attr("y", (d, i) => yScale(i) - height_rect/2) .attr("width", width_rect) .attr("height", height_rect) .attr("fill", d => color(colorScale(d["values"]))) let tooltip = chart.append('g') .attr("class", "tooltip") rectangles.on("mouseover", function(d, i){ render_tooltip(tooltip, xScale(d["expression"]), yScale(i), d["values"]) d3.select(this) .style("stroke", "orange") .style("stroke-width", "1.5") }) .on('mouseout', function(d) { tooltip.selectAll("rect") .remove() tooltip.selectAll("text") .remove() tooltip.style("visibility", "hidden"); d3.select(this) .style("stroke", "orange") .style("stroke-width", "0") }) chart.append("g") .call(xAxis) .selectAll("text") .style("text-anchor", "end") .attr("dx", "-0.9em") .attr("dy", ".15em") .attr("transform", "rotate(-30)") .append("text") .attr("class", "axis-label") .attr("transform", `translate( ${(width*cLeft.w + margin.left*(1 + cLeft.m.l) - margin.right*(1 + cLeft.m.r)) / 2}, ${2 * margin.bottom*(1 + cLeft.m.b) / 3} )`) .attr("font-size", 14) .text("Tokens") chart.append("g") .call(yAxis) .append("text") .attr("class", "axis-label") .attr("transform", 'rotate(-90)') .attr("x", - (height*cLeft.h / 2)) .attr("y", - (2 * margin.left*(1 + cLeft.m.l) / 3)) .attr("font-size", 14) .text("Volumes") chart.append("g") .append("text") .attr("class","chart-title") .attr("transform", `translate(${margin.left*(1 + cLeft.m.l)}, ${2 * margin.top*(1 + cLeft.m.t) / 3})`) .text(title) render_legend_color_range(svg_matrix_chart, context, color, [max_number, 0]) return svg_matrix_chart.node() }38:T557,function show_highlights(data, obj){ let svg = obj["svg"] let xScale = obj["xScale"] let yScale = obj["yScale"] const offset = 3*xScale.bandwidth()/7 let highlights = svg.append("g") .selectAll("g") .data(data) .enter() .append("g") .attr("class", "highlights") .attr("transform", function(d, i) { return "translate(" + (xScale(d["x"][0]) + xScale.bandwidth()/2 - offset) + "," + (yScale(d["y"].slice(-1)[0])) + ")" }) highlights.append("rect") .attr("width", d => xScale(d["x"].slice(-1)[0]) - xScale(d["x"][0]) + 2 * offset) .attr("height", d => yScale(d["y"][0]) - yScale(d["y"].slice(-1)[0])) .attr("fill", "transparent") .attr("stroke-width", 1) .attr("stroke", "black") .on("mouseover", function(d){ let ticks = svg.selectAll(".tick text") ticks.attr("font-weight", function(e){ if (d["x"].includes(e)){ return 800 } else { return 500 } }) }) .on("mouseout", function(d){ let ticks = svg.selectAll(".tick text") ticks.attr("font-weight", 500) }) highlights.append("text") .attr("transform", "translate(0, -10)") .attr("fill", "black") .attr("font-weight", 700) .attr("font-size", 15) .attr("font-family","sans-serif") .text(d => d["label"]) }39:[] 2e:{"id":"4f4ff6b83b956179","slug":"in-search-of-marcel-proust","title":"In Search of Marcel Proust","thumbnail":"41e9ad527a8930890b7fce282d29a0d460dbadf6ac6164611d26df1b5cdeb286","access_level":"public","create_time":"$D2019-06-01T07:24:55.654Z","update_time":"$D2019-09-08T18:17:18.571Z","view_time":null,"first_public_time":"$D2019-06-19T05:33:18.817Z","has_unpublished_edits":null,"fork":false,"stars":21,"starred":false,"owner":{"type":"team","id":"bb3d22f14e970e88","login":"guillaume-lesaine","name":"Guillaume Lesaine","avatar_url":"https://avatars-next.observableusercontent.com/485611ec9c2d43d51f2f7f785b5f6128f45128a4647cf792aaea2f7deb753d30","tier":"starter_2024"},"version":3315,"publish_version":3315,"role":null,"authored":false,"authors":[{"type":"individual","id":"32ba34f10114808b","login":"guillaume-lesaine","name":"Guillaume Lesaine","avatar_url":"https://avatars-next.observableusercontent.com/485611ec9c2d43d51f2f7f785b5f6128f45128a4647cf792aaea2f7deb753d30","tier":null}],"members":[{"type":"individual","id":"32ba34f10114808b","login":"guillaume-lesaine","name":"Guillaume Lesaine","avatar_url":"https://avatars-next.observableusercontent.com/485611ec9c2d43d51f2f7f785b5f6128f45128a4647cf792aaea2f7deb753d30","tier":null,"role":"editor"}],"comments":[],"custom_thumbnail":true,"trashed":false,"lapsed":null,"subscribed":null,"forks":3,"collection":null,"type":"notebook","body":{"titling":"auto","title":"In Search of Marcel Proust","cells":[{"id":0,"value":"md`# In Search of Marcel Proust\n*Texts used in this notebook are from [http://www.gutenberg.org](http://www.gutenberg.org/browse/authors/p) and [https://fr.wikisource.org](https://fr.wikisource.org/wiki/Auteur:Marcel_Proust).*\n*Data processing in Python is avaible on my Github repo [https://github.com/guillaume-lesaine](https://github.com/guillaume-lesaine/vis-proust).*`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":44,"value":"$30","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":22,"value":"macroAnalysis = md`## Macro Analysis`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":872,"value":"md`Marcel Proust's writing raises mixed feelings. Some readers will tell you that Proust makes sentences that are too long and complicated. Others will emphasize its capacity of revealing thoughts you had but could not formulate. Let's see how a macro analysis can help us better understand how special his writing is.`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":877,"value":"md`We start by showing on the left how sentences are distributed depending on the number of words they contain. On the right, you have the average ratio of punctuation marks per sentence.`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":2688,"value":"{\n let container = html`\n

`;\n \n let div_sentences = create_div('length_sentences', {'float': 'left', 'width': '49%'})\n let div_punctuation_sentences = create_div('punctuation_sentences', {'float': 'left', 'width': '49%'})\n\n div_sentences.appendChild(sentences_chart());\n container.appendChild(div_sentences);\n\n div_punctuation_sentences.appendChild(bar_chart());\n container.appendChild(div_punctuation_sentences);\n\n return container;\n}","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":908,"value":"md`Given that a normal French sentence is 15 words long, let's see how Proust writes:\n - **75%** of words in the book belong in sentences with over 29 words.\n - **25%** of the words are contained in 7% of the sentences.\n - The longest sentence in the book is **844 words** long\n\nAs we can see on the punctuation chart on the right, Proust's writing relies heavily on commas with more than **2.5 commas per sentence**. We can also see that he puts a semicolon almost every ten sentence in order to avoid breaking long structures. Today, French people do not even use semicolons. Finally we can see that Proust **does not really use punctuation like ! or ? to give rythme to its sentences**. They mainly appear in dialogues.`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":2315,"value":"md`Let's now see how the length of sentences is distributed over volumes. Indeed, Proust wrote every day during 16 years. We can thus ask ourselves if his writing evolved over time.`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":2857,"value":"$31","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":2320,"value":"md`It is interesting to notice two related things here. The **heart of the text**, volume 3, seems to be **denser than the rest**. Indeed, the begining and the end of the book seem to have longer sentences overall.`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":2330,"value":"md`To conclude this macro analysis, we want to highlight two facts. First, yes figures show that Proust writes long, calm and complex sentences, using commas and semicolons. Second, it seems that the writing evolved over time or depending on the topics. Indeed, the sentences of volume 3 are more concise than the ones of the other volumes.`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":27,"value":"ngramsAnalysis = md`## N-grams Analysis`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":2347,"value":"md`Now that we considered sentences in la Recherche, let's dive into the words. We did an ngrams analysis over the seven volumes for n = 2 and n = 3. This means that we had two ways of breaking sentences. Here is an example:\n\n*\"For a long time I used to go to bed early.\"*\n\n - **n = 2 | ** 2grams = [*\"For a\"*, *\"a long\"*, *\"long time\"*, *\"time I\"*, *\"I used\"*, *\"used to\"*, *\"to go\"*, *\"go to\"*, *\"to bed\"*, *\"bed early\"*]\n - **n = 3 | ** 3grams = [*\"For a long\"*, *\"a long time\"*, *\"long time I\"*, *\"time I used\"*, *\"I used to\"*, *\"used to go\"*, *\"to go to\"*, *\"go to bed*, *\"to bed early\"*]`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":2366,"value":"md`Below is a first chart showing important ngrams over the sentencees of the book. We omitted common short combination existing in the French languages. We only kept ngrams where at least one gram was more than three-characters long.`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3157,"value":"{\n let container = html`

`;\n \n container.appendChild(violin_ngrams2.svg.node());\n\n return container;\n}","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":2132,"value":"viewof button_ngrams2 = {\n const form = html`