{"id":245,"date":"2020-07-01T11:58:57","date_gmt":"2020-07-01T15:58:57","guid":{"rendered":"https:\/\/arriablog22.wpengine.com\/corpus-analysis-a-great-way-to-understand-what-your-nlg-system-needs-to-do\/"},"modified":"2022-03-24T15:26:11","modified_gmt":"2022-03-24T19:26:11","slug":"corpus-analysis-a-great-way-to-understand-what-your-nlg-system-needs-to-do","status":"publish","type":"post","link":"https:\/\/www.arria.com\/blog\/corpus-analysis-a-great-way-to-understand-what-your-nlg-system-needs-to-do\/","title":{"rendered":"Corpus Analysis: A great way to understand what your NLG system needs to do"},"content":{"rendered":"

Practical advice on building NLG systems<\/strong><\/p>\n

One of the biggest challenges in creating NLG systems is mapping the input data to words and phrases in the narrative. If you have some examples of good narratives (written by a person), then I recommend that you start the process of mapping data to words by marking up (annotating) some of the examples, based on where the data comes from. The process is fairly quick, and can give good insight as to challenges and feasibility.<\/p>\n

I suggest you do this by loading the examples into Word (or whatever editor you use), and color-coding text fragments as follows:<\/p>\n

Unchanging (normal) \u2013 text does not change (i.e., this text does not depend on data)<\/p>\n

Direct data (yellow)<\/span> \u2013 input data<\/p>\n

Derived insights (blue)<\/span> \u2013 insights derived from the data<\/p>\n

Not available (purple)<\/span> \u2013 information is not present or derivable from data<\/p>\n

It\u2019s probably easiest to show this with an example: Let\u2019s assume that the NLG system is supposed to generate a summary of an election result. The developers ask a human journalist to write a short summary of an election result, and she produces the following:<\/p>\n

\u201cKirsty Blackman was re-elected as the MP for Aberdeen North, with 54% of the votes. Her margin of victory was higher than in the previous election, perhaps because her main opponent, Ryan Houghton, was suspended from the Conservative Party because of anti-Semitic comments.\u201d<\/p>\n

Let\u2019s also assume that the input data to the system is election data, including historic as well as current elections<\/a>. In this case, we can annotate what the journalist wrote as follows:<\/p>\n

\u201cKirsty Blackman<\/span> was re<\/span>-elected as the MP for Aberdeen North,<\/span> with 54%<\/span> of the votes. Her margin of victory was much higher than in previous elections,<\/span> perhaps because her main opponent, Ryan Houghton, was suspended from the Conservative Party because of anti-Semitic comments.\u201d<\/span><\/p>\n

In this case,<\/p>\n