You must create an account and log in to the NewsEye Platform to use the platform. At the end of this lesson, you will be able to search the platform's newspaper articles, analyze the results and build your own dataset. 

The platform is composed of three main features : 

• the search engine, 
• the dataset management system 
• the experiment system

When you click on the "Search" menu item, you are taken to a page where you can enter textual queries to find documents. 

The examples of queries proposed below are enclosed between backticks like this: `your actual query`. These characters are not required for a real search with the platform. 

There are two ways of searching for information:

• Stemmed search: in this kind of search, the keywords used in the query are replaced by their stemmed form, i.e. their grammatical root (for example, "presidential" will be transformed into "president"). This kind of search allows for a more significant number of results returned, containing words in a form the user was not necessarily aware of. However, it can also create more noise by returning documents not relevant to the research subject.
• Exact search: as its title suggests, this type of search will identify documents containing the keywords exactly how they were expressed in the search bar.

The exact and stemmed searches are two of the possible search features the user can use. More complex search features are available to make a query more precise to find the most relevant documents.

• Phrase search: When you want to query a document for multiple terms appearing next to each other, you can add quotes. For example, the query "washington dc" will match documents containing these two terms next to each other in the same order. The phrase search can be combined with the following search functions.
• Wildcards: `?` matches a single character. For example, `wom?n` will match documents containing "women" or "woman". `*` matches one or several characters. For instance, `balti*` will match documents containing "baltic", "baltique", or "baltikum". This tool may or may not work as it is very expensive to run.
• Fuzzy search: It is based on the Damereau-Levenshtein distance and allows the querying of syntactically similar terms. The maximum default distance is set to 2. For example, the query `roam∼` will match documents containing words like "roam" and "foam" but not "roast". The distance between "roast" and "roam" is 3 (two deletions and one addition). It is possible to set the maximum distance to 1 using the following syntax: `roam∼1`.
• Proximity search: It is sometimes useful to query documents for words appearing close to one another. For example, the query `"woman vote"∼10` will match documents containing these two terms, separated by ten words at most.
• Term boosting: You can affect how the pertinence of documents is computed (and thus the order of the results) by specifying the importance of the terms in the query. For example, the query `president rooseveltˆ2` will affect the score by giving the term "roosevelt" twice as much importance as the term "president". It is also possible to negatively affect the importance of a term by specifying a number between 0 and 1 (for example, `presidentˆ0.5 roosevelt` is equivalent to the previous query).
• Boolean NOT: To query for documents that should not contain a particular term or phrase, you can prepend it with the `-` symbol. For example, the query `president -roosevelt` will return documents containing the term "president" but not the term "roosevelt".

Suppose you are looking for precision, i.e., articles on a particular topic. In that case, it is best to limit the number of results by choosing relevant keywords and excluding terms that may produce noise in the results.

On the contrary, if you are looking for articles in a broad field, it is better to vary the keywords and allow the search engine to be more flexible on how it handles your query. This will lead to a more significant number of results, often at the expense of their accuracy.

In most cases, a combination of both approaches is ideal. In most cases, a combination of both approaches is ideal. It is often necessary to start with a broad search and then gradually refine it by adding or removing keywords or refining them.