text summarization
In plain wordsUsing a computer to shorten a document, meeting or set of articles into a brief version that keeps the main points.
The task of condensing one or more source texts into a much shorter one that keeps what matters most for the reader's purpose. Systems are either extractive, copying out the sentences judged most central and stitching them together, or abstractive, composing new sentences that compress, rephrase and merge material from different parts of the source. Variants include single- and multi-document, query-focused, update, and dialogue or meeting summarization. Early systems scored sentences by word frequency, position and cue phrases; neural sequence-to-sequencesequence-to-sequence model, n.: A model that turns one sequence into another, such as a French sentence into English or speech into text, by first reading the whole input and then writing the output step by step. models made abstractive summarization practical from 2015, and LLMslarge language model, n.: A computer program trained on enormous amounts of text to predict the words that come next. At vast scale, that simple skill lets it chat, answer questions, translate, summarize and write code. now dominate. Evaluation uses ROUGEROUGE, abbr.: An automatic score for summaries that checks how much of the wording in human-written reference summaries also appears in the computer-generated one. overlap with reference summaries, human judgment and model-based judges; the central risk is unfaithfulness — summaries that add or distort facts not in the source (hallucinationhallucination, n.: When an AI states something false or made up — a fake fact, quotation or source — fluently and confidently, as if it were true.).
Abstractive summarization reads better, but check that the figures match the report.
Origin
Hans Peter Luhn of IBM, building on his 1957 statistical work on text, built 'auto-abstracts' from statistically significant sentences in 'The Automatic Creation of Literature Abstracts' (1958). Harold Edmundson (1969) added position and cue-word features; Chin-Yew Lin's ROUGE (2004) became the standard metric, and Alexander Rush, Sumit Chopra and Jason Weston's 2015 neural attention model launched modern abstractive summarization.
Related
- natural language generationnatural language generation, n.: Getting computers to produce human language — from turning a spreadsheet into a weather report to writing an essay or a chatbot's reply.
- ROUGEROUGE, abbr.: An automatic score for summaries that checks how much of the wording in human-written reference summaries also appears in the computer-generated one.
- hallucinationhallucination, n.: When an AI states something false or made up — a fake fact, quotation or source — fluently and confidently, as if it were true.
- factualityfactuality, n.: How far what an AI says is actually true about the world. A factual model gets facts, numbers, names and dates right and does not make things up.
- sequence-to-sequence modelsequence-to-sequence model, n.: A model that turns one sequence into another, such as a French sentence into English or speech into text, by first reading the whole input and then writing the output step by step.
- large language modellarge language model, n.: A computer program trained on enormous amounts of text to predict the words that come next. At vast scale, that simple skill lets it chat, answer questions, translate, summarize and write code.
Sources
- Luhn, H. P. (1958). The Automatic Creation of Literature Abstracts. IBM Journal of Research and Development 2(2), 159–165.
- Lin, C.-Y. (2004). ROUGE: A Package for Automatic Evaluation of Summaries. Text Summarization Branches Out, ACL Workshop. aclanthology.org/W04-1013
Mentioned in
natural language processingnatural language processing, n.: The field that teaches computers to handle human language — to read, interpret, translate, summarize and produce text and speech. It is one of the oldest and largest areas of artificial intelligence., GraphRAGGraphRAG, prop. n.: A way of letting AI answer questions about a big pile of documents by first turning them into a map of people, things and their connections, then using that map to find and summarize answers., natural language generationnatural language generation, n.: Getting computers to produce human language — from turning a spreadsheet into a weather report to writing an essay or a chatbot's reply., ROUGEROUGE, abbr.: An automatic score for summaries that checks how much of the wording in human-written reference summaries also appears in the computer-generated one., sequence-to-sequence modelsequence-to-sequence model, n.: A model that turns one sequence into another, such as a French sentence into English or speech into text, by first reading the whole input and then writing the output step by step.