Practical applications of natural language text analysis systems

Question

Analytics

What would be the use of a natural language text analysis system that could perform a complete semantic analysis of almost any text? Interested in the practical application of such systems (like this one ), tk. most articles and books mention such applications: “analyzing information for target criteria” or “highlighting key information” - admittedly, such “water” does not give any idea of ​​\u200b\u200ba possible application.
Perhaps someone in their projects has needs for such systems - I would like to hear about these “needs” (in detail).

Answer 1

The achievable goals of this task now are:
* autocatalog (stuff the news stream by topic)
* abstract (select from the text - or compose - something that will briefly describe it, and show it when searching)
An unattainable goal for today is to highlight entities, eliminate syntactic and semantic ambiguity, save connections and relationships in the knowledge base (example: learn about a new unit of measurement from the text of this article and be able to convert it to liters). That is, to understand the meaning of the text as a person understands it.
If about my own about practical, then I did not participate in this for longproject. I used it (oh...) to auto-rewrite texts taking into account the context (so that in the text about cars I would replace the word “oil” with “petroleum”, but not with “flattery”).

Answer 2

To find out which candidate will win the election, it is enough to count how many times his name appears in the press. Simply put, how many times people have been brainwashed. You can also learn about the development of a new weapon if the key components of the system began to be discussed more in the press, or vice versa, the topic suddenly disappeared from the press. In the latter case, this happened during the development of a nuclear bomb by the Americans.
If you pay attention to the construction of the text, you can build templates. If the text is deliberately distorted, then what is the reason for this? Is it just young people having fun or are they trying to hide something? With sufficiently powerful servers, you can achieve very high-quality text analysis based on templates.
Based on everything, you should understand why “water” is written and no specifics.

Answer 3

I am currently working on a similar system (the link to which the author provided). Certain categories of entities are defined (full names, names of organizations, car brands, phone numbers and a bunch of other things), rules are written for them (how to highlight them in the text). Based on the selected entities, a semantic network is built. Rules are also written to highlight relationships between entities. Well, the actual resulting network (or graph) is beautifully visualized, if necessary, fits into the base. Unfortunately, I can’t tell you in more detail, but the scope is very wide. A particularly good result is obtained if there is a huge pile of rather similar documents, various kinds of operational information (there are fewer rules for highlighting them).
If somehow I find time and the authorities are not strongly opposed, I will definitely write an article here, I think people will be interested.

Answer 4

Sentiment analysis for large volumes of texts (social media on a global scale). The better the internal text analyzer, the more accurate the reports. I am currently working in this area.

Answer 5

We are thinking about something like this for our project - a person writes in the search, de, "three-room villa in Mallorca at the end of August" - and he hop, shows the search results with exactly the specified parameters.

Answer 6

It's not water at all. I translate:
"analysis of information for target criteria"
means the analysis of advertisements such as buy <-> sell, finding matches and issuing the resulting list. The most urgent task.
"Identification of key information"
means the analysis of texts for the mention of given information in a given context. A crude example of "blow up the embassy." In practice, commercial applications are of much more interest, this is targeted advertising in the first place and semantic search in the first place. And yes, abstracts, but that's later, because before that, the computer has already learned to speak.

Answer 7

A very urgent task for many things is automatic classification. For example, from a long, lengthy description of wrenches, you can understand that they are such and such, wrench number 1 is like this, and wrench number three is different.

Answer 8

Here we are making systems for creating technological foresights (this is something between plans and forecasts). People collect the so-called. signals - events that may be important to the industry, and based on several signals, they should have trend ideas. There are a lot of signals, so I want to show the user small groups of signals that are related or somehow similar. This is where a classifier would come in handy.