MTCB
Collection
Massive Text Chunking Benchmark • 8 items • Updated
id int64 0 999 | text stringlengths 14.7k 98.1k | lang stringclasses 10
values | path stringlengths 5 158 | repository_name stringlengths 8 76 |
|---|---|---|---|---|
0 | // Copyright 2021 Google LLC.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// Code generated file. DO NOT EDIT.
// Package webrisk provides access to the Web Risk API.
//
// For product documentation, see: https://cloud.google.com/web-risk/
//
// Creating a c... | go | webrisk/v1/webrisk-gen.go | nimbix/google-api-go-client |
1 | "<?xml version=\"1.0\" ?><!DOCTYPE TS><TS language=\"ca\" version=\"2.0\">\n<defaultcodec>UTF-8</def(...TRUNCATED) | typescript | src/qt/locale/bitcoin_ca.ts | elifes/coin |
2 | "<?xml version=\"1.0\" ?><!DOCTYPE TS><TS language=\"bs\" version=\"2.0\">\n<defaultcodec>UTF-8</def(...TRUNCATED) | typescript | src/qt/locale/bitcoin_bs.ts | 2pay/MeCoin |
3 | "<?php\n\n/*\n * This file is part of the Sonata package.\n *\n * (c) Thomas Rabaix <thomas.rabaix@s(...TRUNCATED) | php | src/Sonata/Bundle/DemoBundle/DataFixtures/ORM/LoadProductData.php | igormukhingmailcom-forks/sandbox |
4 | "\n/* System module */\n\n/*\nVarious bits of information used by the interpreter are collected in\n(...TRUNCATED) | c | contrib/tools/python3/src/Python/sysmodule.c | AnnaAraslanova/catboost |
5 | "// This file is by code gen, do not modify\n\nimpl<'a, T0: Clone, T1: Clone> TupleCloned for (&'a T(...TRUNCATED) | rust | tuples/src/gen/cloned.rs | BatchOperator/tuples |
6 | "// Code generated by software.amazon.smithy.rust.codegen.smithy-rs. DO NOT EDIT.\n/// Error type fo(...TRUNCATED) | rust | sdk/backupgateway/src/error.rs | StevenBlack/aws-sdk-rust |
7 | "use std::fmt::{self, Write};\nuse std::mem::MaybeUninit;\n\n#[cfg(all(feature = \"server\", feature(...TRUNCATED) | rust | src/proto/h1/role.rs | MarinPostma/hyper |
8 | "###########################################################\n#\n# Copyright (c) 2005, Southpaw Tech(...TRUNCATED) | python | src/tactic/ui/container/tab_wdg.py | CeltonMcGrath/TACTIC |
9 | "# frozen_string_literal: true\n\n# WARNING ABOUT GENERATED CODE\n#\n# This file is generated. See t(...TRUNCATED) | ruby | gems/aws-sdk-cloudsearch/lib/aws-sdk-cloudsearch/client.rb | AllanFly120/aws-sdk-ruby |
Cocha is a dataset of source code files from popular GitHub repositories across multiple programming languages, designed to evaluate how well chunking algorithms handle code with its unique structure and syntax.
This dataset tests chunking algorithms on:
| Field | Description |
|---|---|
id |
Unique identifier |
text |
Full source code content |
lang |
Programming language |
path |
File path in repository |
repository_name |
Source repository |
| Field | Description |
|---|---|
doc_id |
Reference to corpus document |
path |
File path |
lang |
Programming language |
repository_name |
Source repository |
question |
Question about the code |
answer |
Answer to the question |
chunk-must-contain |
Code passage that must be in the retrieved chunk |
from datasets import load_dataset
# Load corpus
corpus = load_dataset("chonkie-ai/cocha", "corpus", split="train")
# Load questions
questions = load_dataset("chonkie-ai/cocha", "questions", split="train")
Cocha is part of the Massive Text Chunking Benchmark (MTCB), a comprehensive benchmark for evaluating RAG chunking strategies.
CC-BY-4.0