What Parsers do

Today, the team offers a browser extension that allows to extract and visualize unstructured data from web pages without programming skills.

It is enough to select the relevant data on one page. The service will automatically find similar ones by a given pattern. Do you need to collect 100 thousand News Banners? Parsers will find them and present in a structured single-fashioned table.

The algorithm can collect links, images, texts, tables and scripts. The structured data in the form of tables or charts can be downloaded in different forms — from the website or by API. An automatic parsing can be configured on a schedule. This may be convenient for comparing the prices, as an example.

The service sells on subscription. The price depends on the amount of collecting information. The free subscription allows to process up to a thousand pages on one site. The most expensive subscription plan allows perfecting 100 thousand pages for the cost of $199 per month.

Microsoft, Cisco, eBay and other big companies use the Parsers product, but the company does not intent upon to stop on a simple parsing on a request. Their agenda is much loftier. Evgeniy Gurinovich, a founder of the company, has a detailed story hereinafter.

...

The new search

We are moving towards creating a new search model when a person receives an answer to his question immediately and does not spend plenty of time on extending the information from the multiple links in a search engine. Practically, all search engines are engaged in parsing. The word parsing is usually replaced by the indexing, still, the essence remains the same. Search engines parse everything, and it is unstructured information.

There is the form of the requests’ result that users get. Browsers give a link to a page where something similar to the query exist. If you ask “How much does the iPhone cost?”, the search engine won’t show an answer on the page, it will send you searching the sites yourself. Search engines do not know where the price is on the site, and especially, if there is any.

They cannot determine what gets in their base. Website structures are so diverse, that it is impossible to use the technical solution to distinguish the price of goods, their weight, color and other parameters. Only a human is capable for it, in spot even not all of them. Yet, it is possible to automate the process partially.

Google and Yandex are far too big to change the search model at short notice. Nevertheless, in small steps, they move to establish a search on structured data and launching SERP (Search Engine Result Page) features as “searchster”, “Goodies”, “instant answers”. Yet, it remains on the basis of the individual sites, not the entire Internet as a whole.

The unmentioned cases show that this approach works. Users feel convenient getting answers immediately, without any extra clicks. That much convenient, that Avito, CIAN, “2ГИС” recently complained about Yandex “searchsters” that take up to half of the overall traffic from them.

We are following the way of creating the structured data output based on the unstructured data collection. We already have a  search engine for companies. This is a beta version. We collected data from the internet on 130 thousand businesses and continue seeding the database now.

For user, it is enough to open one webpage in order to see all the necessary information. It solves the problem of opening dozens of tabs in the browser. Voice assistants also work with structured facts.

...

...

The competitors

Everything connected with the Internet search, intercrosses with big companies’ interests, such as Google and Yandex. However, our archrival is the Diffbot company.

This company collected the database of more than 3 trillion structured facts. At the same time, a similar Knowledge Graph from Google contains one order less. Diffbot focuses on B2B sales, its clients are Cisco, Amazon and U.S. government.

Diffbot has attracted $10 million in round A, being estimated over 100 million. Analogous startups preferred to exit. Software developer Palantir Technologies went off to Kimono Labs and technical giant IBM bought AlchemyAPI.

...

What is next?

At the beginning of July, we launched the companies search. We will keep a close eye on the users’ feedback, expand the data and add new data types. There is just the initial version of the search engine that would work for B2C and look different in the future. Although, even now it is much easier to find some information about the companies. If you are making a list of investors, just choose the Venture category and get the list that may be segmented by countries and other criteria.

We will add the search by product, real estate, events and other sections. After all, all sections will be combined with the simple search engine interface.

As for today, we develop the product for personal resources, negotiate with investors and already have a potential one. It is important for now to choose the investor that will help us improve the USA and East European market positions. For us, it is more important to modify the product that is easy to scale and develop in the future, than to get one-minute profit now.

...

Advice to colleagues

Teams that plan to work with data structuring should understand that they need huge amounts of data. Think through how you will collect, store and process it. You must have a unique solution to at least one of these questions. Otherwise, the project’s economy will not converge.

The high rank on Project Hunt Top-day is closer if you follow certain rules. Most of them are described in details in the relevant recommendations. We add some obvious ones.

1. It is necessary to gather minimum 20–30 Product Hunt users that vote for your project during the first few minutes. These votes provide a hit on the main page.

2. Even if you are not on the main page at once, these votes enable you to contact the Product Hunt Support. They respond quickly and can add you. We did not hit the main page immediately, although there were many votes, so we expected to. A few hours later, I contacted the support and they finally placed us immediately. These few hours cost many voices. Perhaps that is the reason why we did not reach the first place.

3. It is better to refer the links for voting to the main page instead of the product page.

4. Fewer votes from those who passed the test and registered on Product Hunt for a long are better than the voices from newly registered accounts. The second can worsen your position.

5. Few weeks before the launch, vote for other projects in order to add the value to your account.

6. There is a huge and very strong community of Product Hunt users in Belarus. Do not hesitate to ask them for help, they can contribute greatly.

The rest depends on your product, the commentators’ approof and luck. Launching of other cool products on the same day can do bad favor to you