File size: 3,256 Bytes
3fc8e60 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 | import type { CitationView } from "@/lib/api";
/**
* Split answer text into plain segments and citation segments.
*
* The regex mirrors the verifier's on the server so the UI can never render a chip the
* verifier did not adjudicate: every match is looked up in the verified citation list,
* and anything unmatched is rendered as plain text rather than silently styled as a
* confirmed source.
*/
const CITATION_RE = /\[\s*([A-Za-z][A-Za-z .'-]{1,48}?)\s*,\s*(?:Article|Art\.?)\s*\(?\s*(\d{1,3})\s*\)?\s*\]/gi;
export type Segment =
| { kind: "text"; value: string }
| { kind: "citation"; value: string; lawLabel: string; articleNo: number; citation: CitationView | null };
export function segmentAnswer(text: string, citations: CitationView[]): Segment[] {
const byKey = new Map<string, CitationView>();
for (const citation of citations) {
byKey.set(`${normalise(citation.law_label)}|${citation.article_no}`, citation);
}
const segments: Segment[] = [];
let cursor = 0;
CITATION_RE.lastIndex = 0;
let match = CITATION_RE.exec(text);
while (match !== null) {
const [raw, lawLabel = "", articleRaw = "0"] = match;
if (match.index > cursor) {
segments.push({ kind: "text", value: text.slice(cursor, match.index) });
}
const articleNo = Number.parseInt(articleRaw, 10);
segments.push({
kind: "citation",
value: raw,
lawLabel: lawLabel.trim(),
articleNo,
citation: byKey.get(`${normalise(lawLabel)}|${articleNo}`) ?? null,
});
cursor = match.index + raw.length;
match = CITATION_RE.exec(text);
}
if (cursor < text.length) {
segments.push({ kind: "text", value: text.slice(cursor) });
}
return segments;
}
function normalise(label: string): string {
return label.toLowerCase().replace(/[^a-z0-9]+/g, "");
}
/**
* Find the span of `passage` that best matches the answer, for highlighting.
*
* Deliberately conservative: it highlights the longest run of consecutive sentences
* whose distinctive words appear in the answer. Highlighting nothing is better than
* highlighting the wrong clause in a legal document.
*/
export function highlightRanges(passage: string, answer: string): [number, number][] {
const answerWords = new Set(
answer.toLowerCase().match(/[a-z0-9]{4,}/g) ?? [],
);
if (answerWords.size === 0) return [];
const ranges: [number, number][] = [];
const sentenceRe = /[^.;:]+[.;:]?/g;
let match = sentenceRe.exec(passage);
while (match !== null) {
const sentence = match[0];
const words = sentence.toLowerCase().match(/[a-z0-9]{4,}/g) ?? [];
if (words.length >= 4) {
const hits = words.filter((word) => answerWords.has(word)).length;
if (hits / words.length >= 0.6) {
ranges.push([match.index, match.index + sentence.length]);
}
}
match = sentenceRe.exec(passage);
}
return mergeAdjacent(ranges);
}
function mergeAdjacent(ranges: [number, number][]): [number, number][] {
if (ranges.length === 0) return [];
const merged: [number, number][] = [ranges[0]!];
for (const range of ranges.slice(1)) {
const last = merged[merged.length - 1]!;
if (range[0] - last[1] <= 2) last[1] = range[1];
else merged.push(range);
}
return merged;
}
|