Classification Functions

Entity Extraction, Linking, Classification, and Tagging for Social Media: A Wikipedia-Based Approach.

Classification_functions.py - one_node_types

From the Web Information Retrieval's project

In the navbar you can find all the link to the codes with their explanation.

The entities have only one node in common

If the list containing the common nodes has lenght equal to 1, we have to extract the only node in, it will be our representative node.

    representative_node = common_nodes[0]
    types = list(representative_node[1])
    not_in = False
    print(types)

Then we extract the types of tweets written by the current author, in particular the types are not strings, since they are extracted from p, they are tuples (type t, probability_of_t). We sort them in decreasing order. If the type with the highest probability is one among the types contained in the representative node we return it.

    author_types = p.get(author)
    max_type = max(author_types, key=lambda item: item[1])
    possible_type = max_type[0]

    if possible_type in types:
        print("The possible type is " + possible_type)
        return possible_type

We can have one or more types in the list of this single node.

    if len(types) > 1:

If we have more than one type we can safelly remove the “Thing” type that it is not so informative.

        types.remove("Thing")
        select_type = types
        predicted_tag = ""

While the predicted tag is not one of the classes of our data we keep searching through the types of the node. We make this check since in the Google Knowledge Graph there a huge variety of types, so we needed to do a simplification, otherwise the number of classes would grow very fast!

        while predicted_tag not in target_names:
            print(predicted_tag)

            # if the list of types we exit from the while-loop
            if not select_type:
                not_in = True
                break
            predicted_tag = class_adjust(select_type[0])
            select_type.pop(0)

If we don’t have a match with our set of classes we decide for the most general class “Thing”.

        if not_in:
            # in the case we have scanned all types we select the "Thing" as class for the tweet
            predicted_tag = "Thing"

        return predicted_tag

If we have only one type, we have to select it in the case is acceptable, otherwise “Thing” again.

    else:
        if types[0] in target_names:
            predicted_tag = types[0]

        # otherwise we have to choose for "Thing", as in the case above.
        else:
            predicted_tag = "Thing"

    return predicted_tag