36
Post Modern PR Crisis Management A paper on the nature, influences and Public Relations responses to crisis in an internet mediated era of ubiquitous interactive communication. By David Phillips FCIPR About the author David Phillips FCIPR David has been active in Public Relations for four decades and has experience in politics, corporate affairs and Hi-Tec PR both in-house and in agency practice. He founded one of the first PR evaluation companies in the 1980’s and is a director of the online monitoring and evaluation company Klea Global. He lectures at Gloucestershire University and is Head of Digital at Publicasity, the London PR agency. He is well published in evaluation and is author of three books on the internet and its significance to public relations.

powerful tools (download)

  • Upload
    nostrad

  • View
    404

  • Download
    0

Embed Size (px)

DESCRIPTION

 

Citation preview

Page 1: powerful tools (download)

Post Modern PR Crisis ManagementA paper on the nature, influences and Public Relations responses to crisis in an internet mediated era of ubiquitous interactive communication.

By David Phillips FCIPR

About the author

David Phillips FCIPR

David has been active in Public Relations for four decades and has experience in politics, corporate affairs and Hi-Tec PR both in-house and in agency practice.

He founded one of the first PR evaluation companies in the 1980’s and is a director of the online monitoring and evaluation company Klea Global.

He lectures at Gloucestershire University and is Head of Digital at Publicasity, the London PR agency.

He is well published in evaluation and is author of three books on the internet and its significance to public relations.

Page 2: powerful tools (download)

Index

Index................................................................................................................................................................................ 2

About the author.............................................................................................................................................................. 3

Introduction...................................................................................................................................................................... 4

The research team.......................................................................................................................................................... 4

Sample selection methodology.......................................................................................................................................4

Methodology for extracting semantic concepts...............................................................................................................6

The Use of Latent Semantic Indexing.............................................................................................................................9

Visualising the data....................................................................................................................................................... 10

Semantics and values................................................................................................................................................... 11

Examining Values in a Crisis.........................................................................................................................................12

The web site defence.................................................................................................................................................14

Managing the ‘sphere of influence’............................................................................................................................14

This week - Toyota in Crisis..........................................................................................................................................15

Conclusions................................................................................................................................................................... 15

Appendix 1.................................................................................................................................................................... 16

Appendix 2.................................................................................................................................................................... 18

Appendix 3................................................................................................................................................................. 22

Appendix 4.................................................................................................................................................................... 25

Page 3: powerful tools (download)

Figures

Figure 1 Method for retrieving citations...........................................................................................................................5Figure 2 Retrieval system checked against automated results.......................................................................................6Figure 3 The Mediations sample post.............................................................................................................................7Figure 4 The source code of the Mediations post...........................................................................................................7Figure 5 Example of text extraction from Mediations post..............................................................................................8Figure 6 Bayesian filtering for cohesive text....................................................................................................................8Figure 7 Checking semantic software in semantic search engine................................................................................10Figure 8 Visualizing semantic concepts in a large corpus.............................................................................................11Figure 9 Toyota semantic values before issues become evident..................................................................................14Figure 10 Evidence of issues in semantic analysis October 2009................................................................................15

Page 4: powerful tools (download)

Introduction

This paper examines how the public relations industry might understand the values expressed about and organisation using semantic analysis of digital texts.

The approach explicates the methodology used to assemble empirical evidence of semantic values evident in 1000 citations per month for the year to January 2010 and all internet citations in the seven days to 23 February 2010 that mention Toyota.

These data are used to present evidence of the changing nature of an organisation facing issues and identifies the tipping point from issues management to full blown crisis.

The research teamI am grateful to colleagues in this research project some of whom have been on this journey with me for many years. The team involved in developing the software include the Times of India Group’s Head of software, Girish Lakshminarayana, Norman Clements, recently Vice President in charge of evaluation at Cision and the author with help from Bruno Amaral and the support of Klea Global Ltd’ software and Publicasity, the PR agency.

Sample selection methodologyThe sample for use in this research was provided through Google’s API1 scraped by Klea Global software using the single search term ‘Toyota’.

Two searches were performed. The first to return 1000 citations at random for each of the 12 months to January 2010, in total 12,000 citations The second included all citations each day in the seven days to 26 February 2010 which amounted to 189,000,000 citations.

An example of the retrieved citations is provided in fig1

1 Google API details are available here http://code.google.com/ accessed February 2010

Page 5: powerful tools (download)

Figure 1 Method for retrieving citations

To double check the use of the API2 scraping software, natural searches using Google was conducted for five days. See fig 2 for sample of concurrent citation retrieval comparing the two methodologies for raw data acquisition:

2 Klea Global is a company owned By the author, Norman Clements and Girish to develop advanced content analysis services http://www.kleaglobal.com/en/ accessed February 2010

Page 6: powerful tools (download)

Figure 2 Retrieval system checked against automated results

Methodology for extracting semantic concepts.

The methodology for identifying semantic concepts involved an automated sequence of computerised processes.

The methodology has been made available to other researchers to test the software used in this research. The test software is available at this URL http://keeneuk.appspot.com/textanalyse.

To demonstrate the methodology in practice the Philip Young blog, Mediations, has been used for demonstration purposes http://publicsphere.typepad.com/.

Specifically, the post entitled ‘Personal Reputation Management: Review’, dated January 2nd 2010 was selected with the URL http://publicsphere.typepad.com/mediations/2010/01/personal-reputation-management-review.html fig 3.

Page 7: powerful tools (download)

Figure 3 The Mediations sample post

Parsed through the Klea Global software the data collected included the URL and the source code see fig 4:

Figure 4 The source code of the Mediations post

The Klea software removes the HTML coding and if the coding is not well forms removes that too.

The result is all the text, much of which is not needed, important or relevant. See Fig 5.

Page 8: powerful tools (download)

Figure 5 Example of text extraction from Mediations post

Using text extraction software the Klea software creates cohesive text. In effect extracting the headline and content of what can be described as an ‘article’ see fig 6.

Figure 6 Bayesian filtering for cohesive text

While this extracts the text, it is not always the text as seen on screen and to extract this content the Klea software

uses a  Bayesian inference using the Bayes' formula for conditional probability: .

The effectiveness of this methodology can be tested here: http://keeneuk.appspot.com/textanalyse.

For the research corpus, the clean text was then parsed through the bespoke Klea Global Latent Semantic Indexing software package.

Page 9: powerful tools (download)

The Use of Latent Semantic Indexing

There are a number reasons we adopted Latent Semantic Indexing as our primary content analysis methodology. We used it as the basis for identifying the most significant words in the corpus which would lead us most closely, accurately and with the most consistent replication to a human understanding of huge numbers of internet citations.

To paraphrase Wikipedia3: Latent Semantic Indexing (LSI) is an indexing and retrieval method that uses a mathematical technique called Singular Value Decomposition (SVD) to identify patterns in the relationships between the terms and concepts contained in an unstructured corpus. LSI is based on the principle that words that are used in the same contexts tend to have similar meanings.

The principles applied are those explicated by Deerwester et al (1990)4

Using statistical analysis, LSI can discover that documents have words which are often used in the same context. For example, "apple" and "computer" will also have "Mac OS" and are therefore also relevant. The same thing applies with "windows" as an operating system as opposed to an invention for looking through walls. It's all about trying to understand more about the nature and intent of the user query and returning information in context with the user's search, even when they give little clue as to the actual nature of the search. LSI is used by other search engines besides Google including Yahoo and Bing."

Additionally, LSI is robust in recommending concepts from content (and is used in a number of commercial applications for doing this based in the work of Rosenstein, and Lochbaum (Rosenstein, M. and Lochbaum, 2000)5

Klea Global has made a copy of its LSI software available online to allow researchers to replicate the experiments and test the outcomes of this research at http://www.netreputation.co.uk/summariser/getconcepts.php.

Using this software, the semantic concepts that are derived from the texts in the Mediations blog post are as follows:

personal reputation, plain stupidity, reputation management, roy murphy, facebook friends, design creativity, radio dj, personal brand, space design, goldbach, serious danger, halpern, internet work, internet age, reading lists, white space, battleground, talk radio, outset, dating site,

Testing the efficacy of the Klea Global semantic analysis software is reasonably easy. Google uses semantics as part of its algorithm and thus, by searching Google with a few (typically no more than three) of the semantic concepts in order of significance, should yield the original article in the search results as our test shows (fig 7).

3 The Latent Semantic Indexing reference in Wkipedia can be found here http://en.wikipedia.org/wiki/Latent_semantic_indexing accessed February 20104 Deerwester, S., Dumais, S. T., Landauer, T. K., Furnas, G. W. and Harshman, R. A. (1990) - no figures, "Indexing by latent semantic analysis." Journal of the Society for Information Science, 41(6), 391-4075 Rosenstein, M. and Lochbaum, C. (2000) "Recommending from Content: Preliminary Results from an E-Commerce Experiment." In Proceedings of CHI'00: Conference on Human Factors in Computing, The Hague, The Netherlands: ACM. http://www.apparent-wind.com/mbr/papers/chi2000.pdf Accessed February 2010.

Page 10: powerful tools (download)

Figure 7 Checking semantic software in semantic search engine

Using this methodology, the research project is able to extract the semantic concepts in each citation.

This research project assembled all the citations and their texts with the associated URL’s and semantic concepts into a single database.

Visualising the data

As can be imagined, there are a lot of LSI concepts derived from a corpus running into over 15,000 individual items retrieved from the Google index. It is not unusual to identify 20 concepts for each item which would suggest the analysis we were attempting would encounter an estimated 300,000 semantic concepts. Many of these concepts would be the same or similar while others would be nearly unique.

There was a need to be able to represent these data in a form that could be readily understood. The solution was developed by Girish Lakshminarayana in which he represented frequently occurring concepts in large fonts and less frequently occurring concepts in smaller fonts. In addition, in time series analysis he represented newly occurring concepts in lighter colours than more established concepts. He called this representation a ‘Reputation Wall’ (fig 8).

Page 11: powerful tools (download)

Figure 8 Visualizing semantic concepts in a large corpus

Not to be confused with a word count ‘word map’, this visualization of semantic concepts is a technique that helped follow the semantic representation of an organization over time relatively easily.

Semantics and values

The significance of values in PR research is probably best explicated by Amaral in his paper to the Euprera conference in 20106. He notes that the study of values is mainly attributed to Milton Rokeach7, Schwartz and Hofstede8.

6 Amaral, B. 2010 Concepts of Values for Public Relations, Euprera Spring Symposium 2010 http://www.euprera.org/ accessed February 2010.7 Rokeach, M. (2000). Understanding Human Values: Individual and Societal. Free Press. 8 Hofstede, G., & Hofstede, G. J. (2004). Cultures and Organizations: Software of the Mind (2nd ed.). McGraw-Hill.

Page 12: powerful tools (download)

The reason for using semantics as the basis for monitoring news and internet citations about organizations go back to the work of many authors who found that semantic analysis allows researchers to program machines to understand human commands via natural language rather than strict programming protocols (Landauer et al 2007)9 .

Semantics is a post ‘Super Theme’ (Phillips 2002)10 approach to content analysis where the theme is determined by the content of the corpus.

Steyvers and Griffin (2009)11 argue that such a correspondence exists between human memory and internet search, and show that this correspondence leads to both better models of human cognition, and better methods for searching the web. We demonstrated that there is a direct link between semantic concept and retrieval of articles in the example above. Indeed, in more robust and prolonged testing, we found that quite frequently a concept would occur on a Reputation Wall which seemed to bear no relationship with the subject matter. However, using the search term (e. g. Toyota in Google, Yahoo or Bing) and a semantic concept (e.g. riverside), the result would always take us to the provenance of the semantic concept (in this case using the concept evident in Fig 8 two active Toyota dealers in Riverside County California and Rome,Georgia).

Aaron Wall revealed more about both the use and results achieved by Google using Latent Semantic Indexing (LSI) in its search algorithm (Wall 2005)12.

Latenet Semantic Indexing behaves in a way that is very like a human being looking at content as is shown by Yu et al (Yu 2002)13 :

Latent semantic indexing adds an important step to the document indexing process. In addition to recording which keywords a document contains, the method examines the document collection as a whole, to see which other documents contain some of those same words. LSI considers documents that have many words in common to be semantically close, and ones with few words in common to be semantically distant. This simple method correlates surprisingly well with how a human being, looking at content, might classify a document collection. Although the LSI algorithm doesn't understand anything about what the words mean, the patterns it notices can make it seem astonishingly intelligent.

Yu et all also note: One great advantage of LSI is that it is a strictly mathematical approach, with no insight into the meaning of the documents or words it analyzes. This makes it a powerful, generic technique able to index any cohesive document collection in any language.

What we see in the literature is that the LSI mathematical and software combination is able to identify the most significant concepts in any corpus in an uncannily human way. It is the reason why search engines have enhanced the semantic component of their algorithms in recent years to achieve results that are closer to human understanding compared to Boolean and word frequency search on its own.

For our purpose, we wanted to identify concepts in predominantly online content that reflected the values a person might give to an entity. In this sense we were taking a philosophical view of semantic theories that seek to explain

9 Landauer, T K., McNamara, D. S., Dennis, S., and Kintsch, W.(2007) Handbook of Latent Semantic Analysis. Psychology Press10 Phillips, D. 2002 Super-themes for PR evaluation Journal: Journal of Communication Management Volume: 6 Issue: 4 Page: 368 - 38611 Steyvers, M. and Griffiths T. L., (2009) Rational Analysis as a Link between Human Memory and Information Retrieval In N. Chater and M Oaksford (Eds.) The Probabilistic Mind: Prospects from Rational Models of Cognition. Oxford University Press.12 Wall, A. Google Semantically Related Words & Latent Semantic Indexing Technology http://www.seobook.com/archives/000657.shtml accessed February 201013 Yu, C., Cuadrado, J., Ceglowski, M. and Payne, j.s., 2002 Patterns in Unstructured Data Discovery, Aggregation, and Visualization National Institute for Technology and Liberal Education (also available at http://www.seobook.com/lsi/lsa_definition.htm accessed February 2010)

Page 13: powerful tools (download)

phenomena such as, as Dever (2000)14 puts it, ‘truth conditions of and inferential relations among sentences/utterances, anaphoric relations among terms, and ambiguity and incoherence of expressions.” His further explanation of semantic value is a theory in which the semantic value of any complex expression results from the functional application of the semantic value of one of its immediate constituents to the semantic value of the other of its constituents. This is an almost complete description of the working of latent semantic software.

The semantic concepts (words) identified by the software are a close metaphor for values in the semantic versus pragmatic view of Wierzbicka15. They are, after all, words that are derived from exact and near meaning, have relative values (Dever 2000) and are numerically derivative, standing on the shoulders of the content in each citation.

There is academic provenance to suggest that it is not unreasonable to view LSI concept words as personal and institutional values.

Examining Values in a Crisis

One way of approaching the significance of a new approach to public relations data is to examine it under extreme conditions.

We examined the values of an organisation as it faced crisis in 2010. Toyota had been facing a series of issues for sometime, and offered an excellent case study.

In traditional Public Relations theory, the organisation reviews its environment, identifies its objectives, develops strategies and develops them through PR tactics. At each stage it monitors and evaluates progress and outcomes.

To identify the environment for Toyota prior to the development of the issues it faced, we examined the values evident in a sample of 1000 citations drawn from the internet worldwide including data from sites like newspapers, blogs, bulletin boards, social networks, micro blogs and other web pages.

In September 2009, the spread of values identified in 1000 citations selected at random from web pages mentioning Toyota and indexed in September 2009 showed no evidence of an emerging unmanageable issue as shown in fig 9.

14 Dever, J. (2000) “Semantic Value”, forthcoming in the Elsevier Encyclopedia of Language and Linguistics.15 Wierzbicka, A. (1991). Cross-cultural pragmatics: The semantics of human interaction, Trends in linguistics studies and monographs 53. Berlin: Mouton De Gruyter.

Page 14: powerful tools (download)

Figure 9 Toyota semantic values before issues become evident

Within a month, there was a significant change in that a number of concepts had emerged that had the potential to affect the reputation of the company (fig 10).

Page 15: powerful tools (download)

Figure 10 Evidence of issues in semantic analysis October 2009

Once again a sample of 1000 citations mentioning Toyota was selected but this time for new pages indexed in October 2009. The emergence of the concepts ‘accelerator’, ‘defect’ and ‘pedals’ all pointed to an emerging issue.

For years, we have known that “information affects the value chain and, in the network of networks known as the Internet, information management is an important corporate function. The value of information will decline as the volume of it increases, and it will gain added advantage when endorsed by trusted channels. Without valued information, the value of products is low to non-existent. Ethics in the provision, management and protection of information is now an important, if not pivotal, management function16.

As the issues related to the defective accelerator became more evident in values expressed about Toyota, traditional wisdom suggests the company should become more transparent and should face up to the issue.

One way of demonstrating transparency and the commitment to resolving issues is through communication media. For most organisations the most evident media for the widest number of concerned publics is the company’s web sites.

The web site defence

Analysis of new pages indexed each month for the web site www.toyota.com was undertaken.

It showed no evidence of value concepts acknowledging the growing significance of the emerging issues expressed in the semantic values. This suggests that the web site was not seen as a strategic channel for communication.

16 Phillips, D., 2000 Blazing Netshine on the value network: The processes of Internet public relations management Journal of Communication Management, Vol. 5, No. 2, pp. 189 - 206

Page 16: powerful tools (download)

Managing the ‘sphere of influence’

A further Public Relations tactic (Phillips & Young 2009)17 The web pages that linked into the Toyota.com web site in January 2010 is to optimise support in the organisation’s ‘sphere of influence’, most commonly identified as among the organisations and individuals that, of their own volition link to the company web sites.

Once again identified the semantic values evident. We identified third party pages with hyperlinks into Toyota.com pages indexed each month. Here there was considerable evidence of change.

The values associated with the new low emission models and technologies and the Toyota (CSR?) sponsorship activities figured highly until December 2009 and had become very muted by January 2010 (see appendix 3).

This week - Toyota in Crisis

The issues faced by Toyota continued throughout February 2010 and a daily watch has been maintained by the PR agency, Publicasity. The results are shown in Appendix 4.

In the period from the end of January to the third week in February there has been a remarkable change in the values identified. A majority of the concepts are about the issues facing the company. Few values are driven by the company and it can be concluded that the company has lost control of the internet agenda. It is evident from the value concepts analysis that the company is in crisis.

At sometime between the end of January and the 15th of February the company lost control of its reputation.

Further insightsThese findings have a number of applications. We can see from them it is possible to monitor and evaluate the values evident across the internet. The search term ‘Toyota’ had 250,000 citations indexed by Google in the year to October 2009. This is a significant corpus but with modern computing not impossible to monitor and evaluate. It offers opportunities to identify the driving forces behind the growing significance of some concept values and to explore the relationships that cluster round these concepts.

Unsaid here is the changed nature of the type of web sites involved.

In the final week, the mix of web sites hosting citations mentioning Toyota showed a significant proportion of overage in forums compared to blogs and content from the generality of web sites is low compared to both historical analysis and other evaluation conducted about other organisations.

17 Phillips,D., Young, P. (2009) Online Public Relations Kogan Page

Page 17: powerful tools (download)

There are many opportunities to use such data in future research. This paper includes a significant amount of data that can be analysed in a number of ways and there is a case for further research using these data but there it extends much other work.

Conclusions

In the paper Towards Relationship Management, Phillips identifies relationships as being formed and sustained by mutually held and understood values18. In July 2009 Amaral and Phillips presented a paper demonstrating a proof of concept for the application of semantic concepts in identifying value concepts in the formation of social groups among the blogging community19. In this paper we have attempted to demonstrate a form of monitoring and evaluation that can be used in landscaping and monitoring the reputational health and drivers of organisations.

From the evidence we have gathered it would seem that there is a relationship between an organisations’ online presence, the ownership of the online agenda and the organisation’s ability to have some control over its future through its presence online.

This raises but one question. If the organisation had no online presence could it survive, grow and become a leading international brand? If the answer is in the negative, then a lot more work is needed by the public relations industry and its academic following to understand the power of the internet in brand and corporate management.

David Phillips February 2010

Appendix 1

The following screen grabs comprise concept values extracted from 1000 internet citations selected at random and mentioning the word Toyota in each month.

18 Phillips, D. 2006 Towards relationship management: Public relations at the core of organisational development Journal of Communication Management Volume: 10 Issue: 2 Page: 211 – 226

19 Amaral, B. & Phillips, D. 2009 A proof of concept for automated discourse analysis in support of identification of relationship building in blogs. A paper to the Bledcom conference Lake Bled http://www.bledcom.com/home/knowledge accessed February 2010

Page 18: powerful tools (download)
Page 19: powerful tools (download)
Page 20: powerful tools (download)

Appendix 2Semantic values taken from new and changed pages published at Toyota.com.

Page 21: powerful tools (download)
Page 22: powerful tools (download)
Page 23: powerful tools (download)
Page 24: powerful tools (download)

Appendix 3

Semantic concepts taken from pages linking into the Toyota.com web site

Page 25: powerful tools (download)
Page 26: powerful tools (download)
Page 27: powerful tools (download)

Appendix 4

In the week up to 26 February, Publicasity maintained a daily record of newly indexed web pages mentioning Toyota. These included online media coverage, blogs, social networks, Twitter, bulletin boards, Google sidewiki content and other web based pages.

Each of these pages has been analysed for the value concepts evident in the text which are provided here.

Page 28: powerful tools (download)
Page 29: powerful tools (download)
Page 30: powerful tools (download)
Page 31: powerful tools (download)
Page 32: powerful tools (download)
Page 33: powerful tools (download)

Amaral, B. & Phillips, D. 2009 A proof of concept for automated discourse analysis in support of identification of relationship building in blogs. A paper to the Bledcom conference Lake Bled http://www.bledcom.com/home/knowledge accessed February 20 0 Amaral, B. 20 0 Concepts of Values for Public Relations, Euprera Spring Symposium 20 0 http://www.euprera.org/ accessed February 20 0.Deerwester, S., Dumais, S. T., Landauer, T. K., Furnas, G. W. and Harshman, R. A. ( 990) - no figures, "Indexing by latent semantic analysis." Journal of the Society for Information Science, 4 (6), 39 -407Dever, J. (2000) “Semantic Value”, forthcoming in the Elsevier Encyclopedia of Language and Linguistics.

Google API details are available here http://code.google.com/ accessed February 20 0Hofstede, G., & Hofstede, G. J. (2004). Cultures and Organizations: Software of the Mind (2nd ed.). McGraw-Hill. Klea Global is a company owned By the author, Norman Clements and Girish to develop advanced content analysis services http://www.kleaglobal.com/en/ accessed February 20 0 Landauer, T K., McNamara, D. S., Dennis, S., and Kintsch, W.(2007) Handbook of Latent Semantic Analysis. Psychology Press Phillips, D. 2002 Super-themes for PR evaluation Journal: Journal of Communication Management Volume: 6 Issue: 4 Page: 368 - 386Phillips, D. 2006 Towards relationship management: Public relations at the core of organisational development Journal of Communication Management Volume: 0 Issue: 2 Page: 2 – 226Phillips, D., 2000 Blazing Netshine on the value network: The processes of Internet public relations management Journal of Communication Management, Vol. 5, No. 2, pp. 89 - 206

Phillips,D., Young, P. (2009) Online Public Relations Kogan Page

Rokeach, M. (2000). Understanding Human Values: Individual and Societal. Free Press.

Rosenstein, M. and Lochbaum, C. (2000) "Recommending from Content: Preliminary Results from an

Page 34: powerful tools (download)

E-Commerce Experiment." In Proceedings of CHI'00: Conference on Human Factors in Computing, The Hague, The Netherlands: ACM. http://www.apparent-wind.com/mbr/papers/chi2000.pdf Accessed February 20 0.Steyvers, M. and Griffiths T. L., (2009) Rational Analysis as a Link between Human Memory and Information Retrieval In N. Chater and M Oaksford (Eds.) The Probabilistic Mind: Prospects from Rational Models of Cognition. Oxford University Press. The Latent Semantic Indexing reference in Wkipedia can be found here http://en.wikipedia.org/wiki/Latent_semantic_indexing accessed February 20 0Wall, A. Google Semantically Related Words & Latent Semantic Indexing Technology http://www.seobook.com/archives/000657.shtml accessed February 20 0Wierzbicka, A. ( 99 ). Cross-cultural pragmatics: The semantics of human interaction, Trends in linguistics studies and monographs 53. Berlin: Mouton De Gruyter. Yu, C., Cuadrado, J., Ceglowski, M. and Payne, j.s., 2002 Patterns in Unstructured Data Discovery, Aggregation, and Visualization National Institute for Technology and Liberal Education (also available at http://www.seobook.com/lsi/lsa_definition.htm accessed February 20 0)