Introduction

The Internet Movie Database, more commonly known as IMDb, is an online database storing information regarding films, television shows and video games. Such information consists of a cast and crew for each production, fictional characters, biographies, plot summaries, trivia and reviews. IMDb allows registered users to submit new material and edit existing entries. Despite the data being regulated the system has succumbed to occasional errors and misuse. Hence, the task at hand is to develop a system whereby existing data can be used to flag up a new credit if it does not fit in with said existing data. In this case, the project is limited to investigating if the recurrence of actors starring in the same movies together can be used to establish the authenticity of information supplied about a new credit (i.e., whether a certain actor did indeed feature in a movie based on whether they have previously co-starred with any of the other actors in said movie in a previous existing credit) and flag up if this does not fit with previously known credits.

The data chosen to be analysed to uncover this relationship is categorised by actors in relation to movies they have featured in. The structure of the data allows us to analyse the problem as a network, where actors are connected by the movies they have co-starred in. There exists multiple approaches to analyse a network to uncover features of the data, including labelling, clustering, shortest path, PageRank, etc. The basic idea of network analysis independent of approach is to uncover features of the data, then to compare different data points to see how much their features differ.

Task

Through comparing the features of newly submitted data to the features of existing data related to the submission, it is possible to determine if the submission is probable or not, and flag thereafter.

---

Running Andrew's Code works on Windows/Linux/MacOSX

Getting a cleaned up version of the tsv in csv format.

1. Place your tsv file or directory in movie_credits.
2. In moviecredits folder (where the main.py is) run python3 main.py

output

• your actors.list.tsv.csv will be located where the tsv file is.
• A list of top ators with the number of times they worked with their colleagues. (in matrix form)

output Example

[[ 0  0  2  0  0  0  0  0  0]
 [ 0  0  0  0  0  0  0  1  0]
 [ 0  0  0  0  0  0  2  1  0]
 [ 0  0  1  0  0  0  1  0  0]
 [ 0  0  0  0  1  0  0  0  0]
 [ 0  0  0  1  0  0  0  0  0]
 [ 0  0  1  0  0  0  0  0  0]]

Results

lines in the file 18435200

no. of actors 2010940

no. of movies 1042622

Code flow (if you want to understand the code)

1. generate.py
   • create a clean csv with self.filtered_csv
   • make unique actors and unique movies with self.unique_actor_movie
   • make a mapping of actors to movies and movies to actors with self.connection (in id form)
2. connection.py
   • create map actors to actors in related movies (including the weighting) with self.actor2actors
   • Build matrix from actors2actors self._build_matrix
   • Build an adj matrix self._build_adjacency_matrix

movie_credits

Machine learning program to flag new credits

RULES: 1 Movies and recurring TV roles only, no TV guest appearances 2 Please submit entries in the format outlined at the end of the list 3 Feel free to submit new actors

"xxxxx" = a television series "xxxxx" (mini) = a television mini-series [xxxxx] = character name = number to indicate billing position in credits (TV) = TV movie, or made for cable movie (V) = made for video movie (this category does NOT include TV episodes repackaged for video, guest appearances in variety/comedy specials released on video, or self-help/physical fitness videos)

Please look at this in Raw

Observations

• A television series is placed inside a string ""
• we do not need character names which are placed inside []
• (TV) and (V) seem to be still considered movies
  • Documentaries lie in (TV), A documentary is a broad term to describe a non-fiction movie
  • (V) Music feature films lie in (V)
• Do we keep uncredited?
• The first element always seems to be the name of the actor