 |
|---|
| generated using DALL-E |
Proceed to the data story for the results.
The film industry has long been captivated by the allure of remaking classic movies, offering fresh takes on beloved stories to new generations. The motivation behind this project stems from a curiosity about the cyclical nature of storytelling in cinema and the industry's reliance on nostalgia. This project aims to delve deep into the phenomenon of movie remakes by analyzing films spanning decades to uncover patterns that explain why certain movies are chosen for remakes, their differences from their originals, and the factors contributing to their success or failure. By examining genre popularity, revenue, critical reception, and temporal gaps, we seek to tell the story of how and why the film industry reinvents existing narratives. This analysis will not only highlight the evolving tastes of audiences but also provide insights into the decisions made by filmmakers and studios. Perhaps as history repeats itself, so do pivotal movies of the generations.
We briefly outline the research questions that we aim to address in this project. We categorize our questions into two main categories as follows.
What key features or metrics make an original movie more likely to be remade? Explicitly, we are interested in:
-
Whether the genre of the movie, the sentiment of the plot line (which could be determined by inference of LLM model), the time gap between the original and the remake, the popularity, revenue, and critical reception are contributing factors to the likelihood of a movie being remade.
-
How these factors can contribute to a movie being remade more than once?
Overall, we can formalize this section as follows: "What is the difference in the distribution of the movies that are remade and the movies that are not remade?"
In this part, we are interested in comparing the original movie with its remakes. We are explicitly interested in:
-
Comparing the ethnicity and popularity of the cast and the crew, the genre, the critical reception, and the revenue of the movie.
-
The time gap between the original and the remake.
-
The contextual similarities such as historical events may co-occur with the original and remade versions.
To conclude, we can summarise this section as follows: "What is the joint/difference between the distribution of the original movie and the remakes?"
Our project focuses on the CMU Movie Summary Corpus, a dataset that includes information about movies and their characters. However, we require data on movie remakes, and the CMU dataset has to be cleaned and enriched for our task, as described below.
We extracted data from Wikipedia's "List of film remakes" pages (1, 2) and organized it into a tabular format. Afterward, to integrate the crawled data into the base CMU dataset, we retrieve the corresponding WikidataIDs for each movie using its URL.
To address the limitations of the CMU dataset—such as missing values, missing records for remade movies, and its limited feature set—we enhanced it using an offline version of TMDB data from Kaggle. This enrichment process involved:
- Adding missing records related to remade movies.
- Incorporating additional features (e.g., budget).
- Filling gaps in the data using TMDB values.
Additionally, we identified and corrected noisy records in the CMU dataset by cross-referencing them with TMDB, resulting in a more consistent and reliable dataset.
To map the Freebase code to the ethnicities, we get an API key through Google Cloud to access the Google Knowledge Graph to retrieve the ethnicity names and descriptions through their Freebase code.
The methods used for this analysis to be done is including, but not limitied to, data gathering from different sources, and enriching the final data using various sources precisely, advanced feature engineering such as adjusting inflation and extracting meaningful features to account for drifts, extracting sentiments and casts informations as features, implementing advanced casual analysis and statistical regression methods such as propensity score matching to extract the underlying patterns. As mentioned before all of these methods are intended to answer to two research questions, RQ1: What is the difference of movies which are being remade and those which are not, and RQ2: What is the difference of a remake movie with its original reference. We have used these methods where-ever they are applicable.
- Aryan: Gathering and Merging the Data of Remakes, Forming the Ideas, Work toward Casual Analysis and Matchings Algorithms, Visualization, Create GitHub Page, Story Telling Over Data.
- Amirmahdi: Analysing, Visualizing, and Implementing Statistical and Regression Methods to Examine the Difference of Original vs. Remakes.
- Main: Analysing, Visualizing, and Implementing Statistical and Regression Methods to Examine the Difference of Remade vs. Non Remades.
- Maria: Analysing, Visualizing, and Fitting Distributions for EDA on the Data, Sentiment Analysis on the Data.
- Yigit: Enrichment and Cleaning of TMDB/IMDB Data.