Is it possible to implement such an application?

Question

Android

Guys.
In general, I am a beginner developer. I still don’t know much and would like to ask experienced people.
You may think my thought is stupid and impossible, but your opinion is important to me.
The concept of an application for mobile devices was born.
The idea is that a specific request is entered in the application.
Further, on the basis of this request, certain information from sites (reviews) is processed.
And as a result, the application would give the best solution based on reviews.
The main thing is that the result should be conscious and better, i.e. there are many reviews, on the basis of which a single decision is made and displayed as a result at the user's request.
I would like to ask you.
Is it even possible to implement this?
And if so, how? What do I need to study for this and from which side do I approach?

Answer 1

Issued a decision about what exactly? Best solution for what? What are the criteria for this "best solution"? What is not the best solution? What are the criteria here? Conscious result? Is it okay that this requires some kind of consciousness and reason? Today, science cannot even give an unambiguous answer to the question - "What is consciousness?".
In addition, there is such a thing as black PR and purchased reviews. And a huge number of these very reviews on the Internet are fake. And what will be the result after analyzing the fake data? And who will need such a result?
In fact, in this case, you need something like an expert system or just some kind of keyword analytics from a bunch of different informational noise. We load the data there - and it already produces something. The most famous expert system today is IBM Watson. And even produces some results. And it has an API - so you can use it. But the result will depend on the quality of the input data. And the simplest keyword analytics is regular expressions plus counters. This is a few pages of code or less (or more - the code is different).

Answer 2

Try to formulate the following on your own in an adequate form (mathematical)
1. Criteria What is the best solution?
2. Criteria, how to evaluate positive feedback and negative feedback?
3. Criteria, what is the set of reviews?
4. Examples of specific requests?
5. How to collect a sufficient amount of initial data (reviews), how to promote a resource for reviews, how to avoid dishonest cheating?
It is possible to implement, but not profitable. Collecting honest feedback from users is extremely difficult and directly contradicts the commercialization of such an application, since the money is mainly given by the advertiser who wants to promote his product. And to maintain an application and servers and a portal costs money.

Answer 3

Your idea completely copies the usual search engine. Open Google, enter a query and click the "I'm Feeling Lucky!" button.
This is quite difficult to do, especially for a beginner.
If, however, some specific small set of sites is provided for which answers will be searched for, then the task will become easier, but not much.

Answer 4

Judging by the description, this is called a question answering system. You can google info on this phrase.
In the simplest case, something like this can be done through full-text search systems that can create more or less smart indexes based on reviews from the database and then search through them.
But in this case, the system does not generate any new information, but simply returns the most appropriate feedback.

Answer 5

In my opinion, neural networks have nothing to do with it. It is necessary to solve the problem by statistical methods.

Answer 6

1. We make a list of received criteria for each of the reviews (ie, for one).
2. We make a morphological analysis of entities and their relationships for each of the reviews.
3. We analyze the resulting relationship tree for each of the reviews.
4. We get the coefficients of the criteria we are interested in.
5. Based on the totality of weights (or on the basis of our pre-prepared (NS) "tree" of weights) of various criteria for each of the reviews, we determine the weight of our target criterion (cumulative) for each particular review.
6. We sort in descending order all the target weights in accordance with the reviews and display the maximum or their clusters as a percentage (for visual comparison).
7. Profit! and (again) Profit!