“Natural-language generation”的意思、由来-开放百科全书

Natural-language generation (NLG) is one of the tasks of natural language processing that focuses on generating natural language from structured data such as a knowledge base or a logical form. It can be used to produce long documents that summarize or explain the contents of computer databases, for example making news reports (automated journalism), summarizing medical records, generating technical manuals, or generating product descriptions for a large e-commerce site. It can also be used to generate short blurbs of text in interactive conversations (a chatbot) which might even be read out loud by a text-to-speech system.

Automated NLG can be compared to the process humans use when they turn ideas into writing or speech. Psycholinguists prefer the term language production for this process, which can also be described in mathematical terms, or modeled in a computer for psychological research. NLG systems can also be compared to translators of artificial computer languages, such as decompilers or transpilers, which also produce human-readable code generated from an intermediate representation. Human languages tend to be considerably more complex and allow for much more ambiguity and variety of expression than programming languages, which makes NLG more challenging.

NLG may be viewed as the opposite of natural-language understanding: whereas in natural-language understanding, the system needs to disambiguate the input sentence to produce the machine representation language, in NLG the system needs to make decisions about how to put a concept into words. The practical considerations in building NLU vs. NLG systems are not symmetrical. NLU needs to deal with ambiguous or erroneous user input, whereas the ideas the system wants to express through NLG are generally known precisely. NLG needs to choose a specific, self-consistent textual representation from many potential representations, whereas NLU generally tries to produce a single, normalized representation of the idea expressed.^[1]

NLG has existed for a long time{{when|date=March 2019}} but commercial NLG technology has only recently{{when|date=March 2019}} become widely available. NLG techniques range from simple template-based systems like a mail merge that generates form letters, to systems that have a complex understanding of human grammar. NLG can also be accomplished by training a statistical model using machine learning, typically on a large corpus of human-written texts.^[1]

Example

The Pollen Forecast for Scotland system^[2] is a simple example of a simple NLG system that could essentially be a template. This system takes as input six numbers, which give predicted pollen levels in different parts of Scotland. From these numbers, the system generates a short textual summary of pollen levels as its output.

In contrast, the actual forecast (written by a human meteorologist) from this data was:

Comparing these two illustrates some of the choices that NLG systems must make; these are further discussed below.

Stages

The process to generate text can be as simple as keeping a list of canned text that is copied and pasted, possibly linked with some glue text. The results may be satisfactory in simple domains such as horoscope machines or generators of personalised business letters. However, a sophisticated NLG system needs to include stages of planning and merging of information to enable the generation of text that looks natural and does not become repetitive. The typical stages of natural-language generation, as proposed by Dale and Reiter,^[3] are:

For instance, in the pollen example above, deciding whether to explicitly mention that pollen

Document structuring: Overall organisation of the information to convey. For example, deciding to

describe the areas with high pollen levels first, instead of the areas with low pollen levels.

Aggregation: Merging of similar sentences to improve readability and naturalness.

Referring expression generation: Creating referring expressions that identify objects and regions. For example, deciding to use

in the Northern Isles and far northeast of mainland Scotland to refer to a certain region in Scotland.

syntax, morphology, and orthography. For example, using will be for the future

An alternative approach to NLG is to use "end-to-end" machine learning to build a system, without having separate stages as above.^[4] In other words, we build an NLG system by training a machine learning algorithm (often an LSTM) on a large data set of input data and corresponding (human-written) output texts. The end-to-end approach has perhaps been most successful in image captioning,^[5] that is automatically generating a textual caption for an image.

Applications

The popular media has paid the most attention to NLG systems which generate jokes (see computational humor), but from a commercial perspective, the most successful NLG applications

have been data-to-text systems which generate textual summaries of databases and data sets; these

systems usually perform data analysis as well as text generation. Research has shown that textual summaries can be more effective than graphs and other visuals for decision support,^[6]^[7]^[8] and that computer-generated texts can be superior (from the reader's perspective) to human-written texts.^[9]

The first commercial data-to-text systems produced weather forecasts from weather data. The earliest such system to be

deployed was FoG,^[10] which was used by Environment Canada to generate weather forecasts in French and English in the early 1990s. The success of FoG triggered other work, both research and commercial.

Currently there is considerable commercial interest in using NLG to summarise financial and business data. Indeed, Gartner has said that NLG will become a standard feature of 90% of modern BI and analytics platforms.^[12] NLG is also being used commercially in automated journalism, chatbots, generating product descriptions for e-commerce sites, summarising medical records,^[13]^[14] and enhancing accessibility (for example by describing graphs and data sets to blind people^[15]).

An example of an interactive use of NLG is the WYSIWYM framework. It stands for What you see is what you meant and allows users to see and manipulate the continuously rendered view (NLG output) of an underlying formal language document (NLG input), thereby editing the formal language without learning it.

Content generation systems assist human writers and makes writing process more efficient and effective. A content generation tool based on web mining using search engines APIs has been built.^[16] The tool imitates the cut-and-paste writing scenario where a writer forms its content from various search results. Relevance verification is essential to filter out irrelevant search results; it is based on matching the parse tree of a query with the parse trees of candidate answers.^[17] In an alternative approach, a high-level structure of human-authored text is used to automatically build a template for a new topic for automatically written Wikipedia article.^[18]

Several companies have been started since 2009 which build systems that transform data into narrative using NLG and AI techniques. These include Phrasetech, Arria NLG, Automated Insights, Narrative Science, Retresco, Visual NLG and Yseop.

Evaluation

As in other scientific fields, NLG researchers need to test how well their systems, modules, and algorithms work. This is called evaluation. There are three basic techniques for evaluating NLG systems:

An ultimate goal is how useful NLG systems are at helping people, which is the first of the above techniques. However, task-based evaluations are time-consuming and expensive, and can be difficult to carry out (especially if they require subjects with specialised expertise, such as doctors). Hence (as in other areas of NLP) task-based evaluations are the exception, not the norm.

Recently researchers are assessing how well human-ratings and metrics correlate with (predict) task-based evaluations. Work is being conducted in the context of Generation Challenges^[19] shared-task events. Initial results suggest that human ratings are much better than metrics in this regard. In other words, human ratings usually do predict task-effectiveness at least to some degree (although there are exceptions), while ratings produced by metrics often do not predict task-effectiveness well. These results are preliminary. In any case, human ratings are the most popular evaluation technique in NLG; this is contrast to machine translation, where metrics are widely used.

See also

References

1. ^{{cite journal |vauthors=Perera R, Nand P |title=Recent Advances in Natural Language Generation: A Survey and Classification of the Empirical Literature |journal=Computing and Informatics |volume=36 |pages=1–32 |year=2017 |issue=1 |url=http://www.cai.sk/ojs/index.php/cai/article/view/2017_1_1}}
2. ^R Turner, S Sripada, E Reiter, I Davy (2006). Generating Spatio-Temporal Descriptions in Pollen Forecasts. Proceedings of EACL06
3. ^¹{{cite book |title=Building natural language generation systems |publisher=Cambridge University Press |location=Cambridge, U.K. |isbn=978-0-521-02451-8 |last1=Dale |first1=Robert |last2=Reiter |first2= Ehud |year=2000}}
4. ^{{Cite web | url=http://www.macs.hw.ac.uk/InteractionLab/E2E/ | title=E2E NLG Challenge}}
5. ^{{Cite web | url=https://www.kaggle.com/c/datalabcup-image-caption | title=DataLabCup: Image Caption}}
6. ^{{cite journal |vauthors=Law A, Freer Y, Hunter J, Logie R, McIntosh N, Quinn J |title=A Comparison of Graphical and Textual Presentations of Time Series Data to Support Medical Decision Making in the Neonatal Intensive Care Unit |journal=Journal of Clinical Monitoring and Computing |volume=19 |pages=183–94 |year=2005 |doi=10.1007/s10877-005-0879-3 |pmid=16244840 |issue=3}}
7. ^{{cite journal |vauthors=Gkatzia D, Lemon O, Reiser V|title=Data-to-Text Generation Improves Decision-Making Under Uncertainty|journal=IEEE Computational Intelligence Magazine|volume=12|issue=3|pages=10–17|year=2017|doi=10.1109/MCI.2017.2708998|url=https://napier-surface.worktribe.com/687579/1/gui-journal-paper-27Apr.pdf}}
8. ^{{Cite web | url=https://ehudreiter.com/2016/12/26/text-or-graphics/ | title=Text or Graphics?| date=2016-12-26}}
9. ^{{cite journal |vauthors=Reiter E, Sripada S, Hunter J, Yu J, Davy I |title=Choosing Words in Computer-Generated Weather Forecasts |journal=Artificial Intelligence |volume=167 |issue= |pages=137–69 |year=2005 |doi=10.1016/j.artint.2005.06.006}}
10. ^{{cite journal |vauthors=Goldberg E, Driedger N, Kittredge R |title=Using Natural-Language Processing to Produce Weather Forecasts |journal=IEEE Expert |volume=9 |pages=45–53 |year=1994 |doi= 10.1109/64.294135 |issue= 2}}
11. ^S Sripada, N Burnett, R Turner, J Mastin, D Evans(2014). Generating A Case Study: NLG meeting Weather Industry Demand for Quality and Quantity of Textual Weather Forecasts. Proceedings of INLG 2014
12. ^{{Cite web | url=https://www.gartner.com/smarterwithgartner/nueral-networks-and-modern-bi-platforms-will-evolve-data-and-analytics/ | title=Neural Networks and Modern BI Platforms Will Evolve Data and Analytics}}
13. ^{{cite conference |author=Harris MD |title=Building a Large-Scale Commercial NLG System for an EMR |booktitle=Proceedings of the Fifth International Natural Language Generation Conference |pages=157–60 |publisher= |year=2008 |location= |url=http://www.aclweb.org/anthology/W08-1120.pdf }}
14. ^¹{{cite journal |vauthors=Portet F, Reiter E, Gatt A, Hunter J, Sripada S, Freer Y, Sykes C |title=Automatic Generation of Textual Summaries from Neonatal Intensive Care Data |journal=Artificial Intelligence |volume=173 |pages=789–816 |year=2009 |doi=10.1016/j.artint.2008.12.002 |issue=7–8}}
15. ^http://www.inf.udec.cl/~leo/iGraph.html
16. ^{{cite book |author=Galitsky, Boris |title=A Web Mining Tool for Assistance with Creative Writing |journal=Advances in Information Retrieval. Lecture Notes in Computer Science |volume=7814 |pages=828–831 |year=2013 |doi=10.1007/978-3-642-36973-5_95|series=Lecture Notes in Computer Science |isbn=978-3-642-36972-8 }}
17. ^{{cite journal |vauthors=Galitsky B, de la Rosa JL, Dobrocsi G |title=Inferring the semantic properties of sentences by mining syntactic parse trees |journal=Data & Knowledge Engineering |volume=81-82 |pages=21–45 |year=2012 | doi=10.1016/j.datak.2012.07.003}}
18. ^{{cite journal |author1=Sauper, Christina |author2=Barzilay, Regina |lastauthoramp=yes |title=Automatically Generating Wikipedia Articles: A Structure-Aware Approach |journal=Proceedings of ACL| year=2009}}
19. ^Generation Challenges 2009