Connect with us

Big Data

Customer engagement strategies that heed power of social media

Published

on

The power of social media to alter customer engagement strategies — for a product rollout, an executive hire, a policy change — is impressive.

Case in point: Legendary Entertainment did not anticipate the kind of response it would get when it released its trailer for The Great Wall. The decision to cast Matt Damon as the hero in a film centered around the iconic Chinese landmark drew immediate criticism on Twitter and Facebook as another example of a white savior narrative and of whitewashing.

“The whole storyline was meant to be about someone coming into a new culture and learning and growing in that culture,” Matt Marolda, chief analytics officer at Legendary, said at the recent HUBweek, an arts, science and innovation festival in Boston. “But the perception was not that.”

Social media platforms and the swift judgment of the internet are forcing companies to engage in ways they’ve never had to before. And executives from Legendary and Microsoft are sharing their experiences with the new tools for — and rules for — customer engagement.

New tools of engagement

On paper, The Great Wall made sense, according to Marolda. It was 2016, and the U.S. and China were the two biggest movie markets in the world; the East-meets-West film reflected Legendary’s sale to Wanda Group, a massive entertainment company in China. And, based on an analysis Marolda and his applied analytics team did, Damon had an active following and a reputation for taking on high-quality projects.

But what looked good on paper did not translate well to audiences — especially those in the U.S. Marolda said the company reacted to the criticism quickly. For example, the company released a statement from Zhang Yimou, the film’s director whom Marolda characterized as “the Steven Spielberg of China,” defending the casting decision.

After that, the team stood still and observed. “We had time on our side,” said Marolda, adding that the film wasn’t scheduled to be released for nine months. “We could see analytically that the best thing to do was nothing.”

The public ire did cool, but the film couldn’t completely escape the negative press it had received, according to Marolda. The company ultimately decided to shift its marketing strategy. “We then realized that emphasizing the movie’s possibilities outside of the U.S. was as important as emphasizing the movie’s possibilities inside the U.S.,” he said.

The decision appears to have been a good one. While the film bombed in the U.S., it was moderately successful worldwide, and has helped spark a larger conversation about how to make blockbuster films for a global market.

Executives onstage at HUBweek.
Matt Marolda, chief analytics officer at Legendary Entertainment, and Kathleen Kennedy, director of special projects at the MIT Sloan School of Management, onstage at HUBweek.

Customer engagement strategies: Ask three questions

How do companies develop customer engagement strategies that acknowledge the power of social media? A reactive approach — no matter how swift the response or how successful in the short term — doesn’t cut it.   

Brad Smith, president and chief legal officer at Microsoft, talked about the role companies should play in the public discourse and stressed that companies need a moral compass today.

“You have to know the issues for which you’re going to take a stand. And you have to be grounded in a certain set of principles,” he said during a fireside chat at HUBweek with Adi Ignatius, the editor in chief of the Harvard Business Review.

Before weighing in on a controversial issue, Smith suggested that companies ask three questions. First, is the issue important to the business? Smith described this question as “an easy space,” and can include tax law or intellectual property law — topics companies have always weighed in on.

Second, is the issue important to its customers? As data has moved to the cloud, companies have entered into a new kind of relationship with their customers, according to Smith. He said it’s vital that they think about the security and protection and actively take a stand on issues like surveillance and privacy.

Third, is the issue important to employees? The company believes a safe work environment doesn’t automatically equate to employee success. Employees could be hindered by issues outside of the office such as an inability to buy the home they want to buy, get the kind of healthcare coverage they need, or marry the person they want to marry, according to Smith.

So when a bill in North Carolina looked like it would restrict LGBT rights, Smith said it “was not a difficult decision” for Microsoft to voice its opposition. The company has a pretty significant presence in Charlotte, employing about 1,000 people there, and Smith said the issue was “important for our employees outside of the workplace.”

In an effort to be as effective as possible and preserve its relationship with the community, Microsoft will often seek out a local business community — a trusted organization that uses its voice to speak up on issues such as these — to partner with. “I prefer a course that’s going to maximize our chances of being effective and not just maximize our chances of being seen,” he said.

Source: https://searchcio.techtarget.com/news/252451024/Customer-engagement-strategies-that-heed-power-of-social-media

Big Data

A Graph-based Text Similarity Method with Named Entity Information in NLP

Published

on

A Graph-based Text Similarity Method with Named Entity Information in NLP

In this article, the author summarizes the 2017 paper “A Graph-based Text Similarity Measure That Employs Named Entity Information” as per their understanding. Better understand the concepts by reading along.


By Prakhar Mishra, Research Scholar at IIIT-Bangalore

In this blog, I have tried summarizing the paper A Graph-based Text Similarity Measure That Employs Named Entity Information as per my understanding. Please feel free to comment your thoughts on the same!

Problem Statement

 
The authors’ propose a novel technique for calculating Text similarity based on Named Entity enriched Graph representation of text documents. Objectively you can think of this as — Given two documents (D1, D2) we wish to return a similarity score (s) between them, where {s ∈ R|0 ≤ s ≤ 1} indicating the strength of similarity. 1 being exactly similar and 0 being dissimilar.

Proposed Method

 


A Graph-based Text Similarity Method with Named Entity Information in NLP | Pipeline
Proposed pipeline | Image from Source

 

Authors’ propose a set of similarity measures over the n-gram graph representation for text documents. To do so, they propose a 3-step pipeline —

  • Information Extraction — This is a first in the pipeline where they extract relevant information chunks from the text document for which they employ two methods: 1. Extraction of Named Entities 2. Extraction of Top-ranked terms using TF-IDF.
  • Graph Representation — The information extracted from the first step is hashed (to get single node representation for multi-word terms) and used as unique nodes in the graph, whereas all the remaining words are replaced with a single placeholder word. Now, this is a modelling choice or you can think of it as a trade-off parameter to how many placeholder nodes would you want to represent. As the use of a single placeholder word results in a word graph having only one node for all the non-important words, which significantly reduces the size of the n-gram graph and the complexity of similarity operators. Let’s take an example to understand this — For instance, if the input sentence is “My name is Prakhar Mishra. I am a developer”. The pre-processed sentence representation becomes “A A A 213aaeb1 A A A _DEVELOPER”, where, is the placeholder symbol for unimportant words, 213aaeb1 is the hash for Prakhar Mishra and _DEVELOPER is the hash for the word developer. Refer to the below figure to understand this visually —


N-gram graph representation of text example
N-gram Graph Representation

 

The edges are weights that you see in the above n-gram graph are decided based on the co-occurrence count of terms in a sliding window of size L traversing over the pre-processed sentence representation.

  • Graph Similarity Measures — Once we have the graph ready, the authors’ employ metrics such as Value SimilaritySize Similarity and Normalized Value Similarity for measuring the similarity between the two graphs, where,

— Value Similarity: This takes into account the set of common edges between two graphs along with their respective weights. It is mathematically represented as:


value similarity text graphs
value similarity

 

where, e is the common edge between two graphs Gi, Gj and VR(e) is calculated as:



VR calculation

 

— Size Similarity: It takes into account the size of the graphs, which is calculated as:


size similarity measure
size similarity

 

— Normalized Value Similarity: This similarity measure ignores the relative size of the graph during comparison. And is defined as:


normalized value similarity text graphs
normalized value similarity

 

If SS (Size similarity)=0, then value of NVS is also set to Zero.


Depending on the use case one can decide on how to use the above set of similarity measures. We can merge the scores from all the above methods using some pooling function and represent it as an aggregated similarity score. Also, another way is to represent the graph as a vector of similarity scores from the above methods and then perform clustering or classification on-top.


Possible Extensions (My thoughts)

 
We can have a little controlled way of hashing wherein the same hash is given to the same entity groups. As this would inducing categorical similarity in the graph and would also reduce the space/time complexity.


You can also checkout other research paper explanations that i have written —

10 Popular Keyword Extraction Algorithms in NLP

BERT-QE: Contextualized Query Expansion

Beyond Accuracy: Behavioral Testing of NLP Models using CheckList

BERT for Extractive Text Summarization

Automatic Hypernym Relation extraction from Text using ML


Feel free to read the paper and say “Hi” to the authors and appreciate their contribution.


Paper Title: A Graph-based Text Similarity Measure That Employs Named Entity Information

Paper Link: Access Paper

Authors: Leonidas TsekourasIraklis VarlamisGeorge Giannakopoulos


Thank you!

 
Bio: Prakhar Mishra Prakhar is currently a MS (by research) grad student in Data Science at IIIT Bangalore. His research interest include Natural Language Understanding and Generation, Information Retrieval, Unsupervised Machine Learning and Reinforcement Learning.

Original. Reposted with permission.

Related:

Coinsmart. Beste Bitcoin-Börse in Europa
Source: https://www.kdnuggets.com/2021/06/graph-based-text-similarity-method-named-entity-information-nlp.html

Continue Reading

Big Data

4 Most Popular ways of Deployment in Cloud Computing

Published

on



Deployment in Cloud Computing | 4 ways of Cloud Computing Deployment





















Learn everything about Analytics


Coinsmart. Beste Bitcoin-Börse in Europa
Source: https://www.analyticsvidhya.com/blog/2021/06/4-most-popular-ways-of-deployment-in-cloud-computing/

Continue Reading

Big Data

Descriptive statistics | A Beginners Guide!

Published

on



Descriptive statistics | A Beginners Guide! – Analytics Vidhya





















Learn everything about Analytics


Coinsmart. Beste Bitcoin-Börse in Europa
Source: https://www.analyticsvidhya.com/blog/2021/06/descriptive-statistics-a-beginners-guide/

Continue Reading

Big Data

NLP Application: Named Entity Recognition (NER) in Python with Spacy

Published

on



Named Entity Recognition (NER) in Python with Spacy





















Learn everything about Analytics


Coinsmart. Beste Bitcoin-Börse in Europa
Source: https://www.analyticsvidhya.com/blog/2021/06/nlp-application-named-entity-recognition-ner-in-python-with-spacy/

Continue Reading
Aviation3 days ago

Delta Air Lines Flight Diverts To Oklahoma Over Unruly Off-Duty Flight Attendant

Esports4 days ago

Lost Ark Founders Pack: Everything You Need to Know

Aviation3 days ago

Spirit Airlines Just Made The Best Argument For Lifting LaGuardia’s Perimeter Rule

Cyber Security5 days ago

Data Breach that Impacted Both Audi of America and Volkswagen of America

Cleantech5 days ago

Tesla Model S 420 Plaid Is The Best Car In The World (But Not For Me)

Energy5 days ago

Industrial robots market in the automotive industry | $ 3.97 billion growth expected during 2021-2025 | 17000+ Technavio Research Reports

Energy4 days ago

Daiki Axis Co., Ltd. (4245, First Section, Tokyo Stock Exchange) Overview of Operating Performance for the First Three Months Ended March 31, 2021

Blockchain5 days ago

Blockchain technology can help to protect sensitive information

Esports1 day ago

Clash of Clans June 2021 Update patch notes

Blockchain2 days ago

Africa Leading Bitcoin P2P Trading Volume Growth in 2021

Blockchain5 days ago

State-Chartered Banks in Texas to Provide Crypto Custody

AR/VR5 days ago

Facebook Just Bought Its Fifth VR Studio, ‘POPULATION: ONE’ Developer BigBox VR

Blockchain5 days ago

DCR Technical Analysis: Look for Support Levels of $130.13 and $126.01

Esports4 days ago

Genshin Impact Murals: Location Guide

Esports1 day ago

Legends of Runeterra Patch 2.10.0 brings bug fixes and Pool Party cosmetics

Gaming3 days ago

Forza Horizon 5 Announced, Launches November 9

Cyber Security5 days ago

Cybersecurity Degrees in Colorado

Energy5 days ago

Industrial synchronous belt drives market|$ 518.13 million growth expected during 2021-2025|17000+ Technavio Research Reports

Aviation2 days ago

Boeing 727 Set To Be Turned Into Luxury Hotel Experience

Aviation5 days ago

United Airlines Will Not Furlough Flight Attendants This Fall

Trending