Get a GOOD domain name, one that matches your brand. Too bad I didn't follow that advice when first starting out with one of my current websites. If I had, I would have saved myself over $3,500 (USD) Soon more people were reading the articles directly on the blog than through the newsletter. I had timed the market well in this instance. Business blogs were just starting.
We use LinkedIn to ensure that our users are real professionals who contribute and share reliable content. When you sign in with LinkedIn, you are granting elearningindustry.com access to your LinkedIn account, which is used to authenticate you without you having to enter a different user name and password.
We also use this access to retrieve the following information: • Your full name. • Your primary email address. You can revoke this access at any time through your LinkedIn account. We use LinkedIn to ensure that our users are real professionals who contribute and share reliable content. When you sign in with LinkedIn, you are granting elearningindustry.com access to your LinkedIn account, which is used to authenticate you without you having to enter a different user name and password.
We also use this access to retrieve the following information: • Your full name. • Your primary email address. You can revoke this access at any time through your LinkedIn account.
Matching test questions present learners with two items separated into two columns and ask them to match items from the first column to the corresponding items in the second.
The number of items in the first column does not necessarily have to match that in the second - it is totally possible to have more items in the second column than in the first one. Matching Test Questions Advantages And Disadvantages Matching questions are at their best when you need to assess the knowledge gained from a course that features a lot of dates, names, places, and events.
As a rule, with matching test questions, learners get partial credit for answers that are only partially correct. Here are their benefits and drawbacks: Advantages • Best “the amount of course material covered to the amount of time spent to construct the questions” ratio. • Allow for great flexibility and accuracy in counting the learners’ scores.
• Give an objective assessment of the learners’ knowledge. • At their most useful when used in areas mostly dealing with facts. • Least chance of guessing the correct answer compared to other question types.
Drawbacks • Ill-suited for gauging the learners’ higher understanding (analysis and synthesis levels). • Answering matching questions is time-consuming for learners. • Introducing too many options can make it so that the question tests the ability to search first, relevant knowledge second. As you can see from the lists above, matching questions will serve you fine in tests that cover large amounts of material.
Just a few well-constructed matching questions will adequately test the understanding of a single topic, or even of a number of topics at once. They are also remarkably easy to construct: in essence, all you need to do is to is pick the key facts and create lists.
Read further to learn how to do it the right way. Best Practices For Constructing Matching Test Questions • Keep questions short and straightforward. Avoid unnecessary words. • Do not get carried away adding additional items. Having 10-12 items between both columns (5-6 “question - answer” pairs) is the sweet spot. • It is best to arrange the items in the left column according to some criterion (alphabetically, chronologically, etc).
• Make sure that no items in the right column can be matched to more than one item in the left one. However, having an item in the left column serves as the key for more than one item in the right column is all right.
• Avoid positioning matching test questions in such a way that the list is separated in two by a page break. Learners should not have to go back and forth trying to match questions on one page to answers on the other. • When constructing answers, try to keep them interconnected by theme and the manner of presentation. You can find examples of a correctly and incorrectly constructed matching questions below.
Incorrect 1. The year of New York’s founding. 2. The capital of the United States. 3. First president of the United States.
4. The date the declaration of independence was signed. 5. The name of the United States currency. A. 4 July 1776 B.
George Washington C. 1653 D. United States dollar E. Washington Correct 1. The largest planet in the Solar System. 2. The planet humans first landed on. 3. The furthest planet from Earth. 4. The planet with an observable ring system. 5. The smallest planet in the Solar System. A. Mercury B. Neptune C. The Moon D. Jupiter E. Saturn Matching Test Questions With Keylists There is a distinct variety of matching questions that makes use of so-called keylists.
Such questions feature a relatively short list of key elements (3-4) and a much larger one containing possible answers (10-12). Learners are asked to match every answer from the second column to one of the keys in the first. Below is an example of a matching question with keylist. 1. Google 2. Microsoft 3. Apple A. Which of the companies on the list derives its profit primarily from context advertising?
B. Which of the companies on the list first produced fonts for an operating system? C. Which of the companies on the list first proposed using a computer mouse? D. Which of the companies on the list first introduced a graphical user interface in their operating system?
E. Which of the companies on the list first created a browser-based cloud operating system? Classification Questions These are very similar to matching questions with keylists, the only difference being that the learners are asked to sort answers from the second column into groups belonging to separate classes or categories specified in the first column.
Classification questions consist of a description of the task the learner has to perform, the list of elements to be sorted, and the list of categories they have to be sorted into. Below is an example of a classification question. Sort the following animals according to the species they belong to: 1. Mammals 2. Birds 3. Fish A. Whale B. Duck C. Dolphin D. Pelican E. Salamander Matching test questions’ biggest advantage is that they allow you to cover large areas of material without having to spend much time or effort on constructing the questions.
In addition, learners find them easy to read and comprehend. Keep these qualities in mind, and you will surely find matching questions handy.
best matching making through names - An Overview of Fuzzy Name Matching Techniques
Methods of name matching and their respective strengths and weaknesses In a structured database, names are often treated the same as metadata for some other field like an email, phone number, or an ID number. But what happens if you only have a name to lookup a record? This happens quite frequently since humans tend to prefer names to numbers and laws may prevent ID numbers from being created or shared. When names are your only unifying data point, correctly matching similar names takes on a greater importance, however their variability and complexity make name matching a uniquely challenging task.
Nicknames, translation errors, multiple spellings of the same name, and more all can result in missed matches. While there is an abundance of search tools on the market, name search is a different animal than document search, and requires a fundamentally different approach. Different name matching methods are best suited to solve different name matching challenges.
There are many ways to match names, but no one universal solution. The best name matching software uses a hybrid of multiple methods to address the maximum number of name variations: Common key method Pros: Fast execution, high recall Cons: Mostly limited to Latin-based languages; transliterating non-Latin names reduces precision These methods reduce names to a key or code based on their English pronunciation, such that similar sounding names share the same key.
A well-known common key method is , patented in 1918. For example, Cyndi, Canada, Candy, Canty, Chant, Condie share the code C530. Many methods take a similar approach to Soundex, including Metaphone and Double Metaphone. These methods use phonetic algorithms which turn similar sounding names into the same key, thus identifying similar names. Metaphone expands on Soundex with a wider set of English pronunciation rules and allowing for varying lengths of keys, whereas Soundex uses a fixed-length key.
Double Metaphone further refines the matching by returning both a “primary” and “secondary” code for each name, allowing for greater ambiguity. In addition, instead of being tied to English pronunciation of characters, it attempts to encompass pronunciations of other origins such as Slavic, Germanic, Celtic, Greek, French, Italian, Spanish, and Chinese.
For example, Double Metaphone encodes “Smith” with a primary code of SM0 and a secondary code of XMT, while it tags “Schmidt” with a primary code of XMT and a secondary code of SMT. That the names share a primary and secondary code of XMT indicates a degree of similarity between the names which Soundex perhaps overstates and which Metaphone misses. Name Metaphone Key Smith SM0 Schmidt SXMTT While the common key method is fast to execute and has good recall, the precision suffers.
Manual inspection of a few names reveals the precision issues. These names share the Soundex key H245: Haugland, Hagelin, Haslam, Heislen, Heslin, Hicklin, Highland, Hoagland. Metaphone does a better job than Soundex, encoding the above names with different codes except for the very similar pairs Haugland/Hoagland and Heislen/Heslin.
Name Metaphone Key Haugland HKLNT Hagelin HJLN Haslam HSLM Heislen HSLN Heslin HSLN Hicklin HKLN Highland HFLNT Hoagland HKLNT For cases where name similarity is being scored against pairs of names in different scripts—for example Korean hangul vs. English—the name must first be converted to Latin characters, which potentially introduces more errors to the comparison. Particularly in languages such as Japanese where one character can have more than one correct pronunciations, converting first to the Latin script can introduce fatal mistakes.
The common Japanese female name 洋子 can be correctly pronounced Yoko or Hiroko. Transliteration of names (a mapping of characters or sounds in one script to another) produces many possible variations since sounds in one language have to be approximated. Variations introduced by transliteration increases the complexity of the already difficult task of matching names. If الرشید عبد is being evaluated against Abdal-Rachid, but the transliteration of الرشید عبد produces Ar-Rashid, will the names come back as a match—as they should?
Name Soundex Key Metaphone Key Abdal-Rachid A134 ABTLRXT Ar-Rashid A623 ARRXT One common key method, the Beider-Morse Phonetic Matching algorithm, does accept Russian in Cyrillic script and Hebrew in Hebrew script, but is otherwise Latin-bound. List method Pros: Easy to maintain Cons: Computationally intensive (read: expensive hardware needed to run against long lists of names quickly); Cannot handle names the system doesn’t know about; Cannot handle names with missing/added spaces between components; Cannot handle names split between different fields This method attempts to list all possible spelling variations of each name component and then looks for matching names from these lists of name variations.
For example: One system produced 3,024 possible transliterations of this Arabic name “الرشید عبد“ since each separate name component alone has several variations. Here are the first five and last five variations. 1. Abdal-rashid 2. Abdal-rashide 3. Abdal-rasheed 4. Abdal-rashiyd 5. Abdal-rachid … 3020. ‘Abd-errshiyd 3021. ‘Abd-errchid 3022.
‘Abd-errchide 3023. ‘Abd-errcheed 3024. ‘abd-errchiyd Trying to generate every possible name variation has a couple of obvious drawbacks. Name variations which are not in the list will not be found as matches, and perhaps an even greater issue is that of speed and size.
Since multi-part names–particularly non-English names–generate an exponentially growing list of variations, searching through these lists takes time. Given a name with just three components and 20 possible variations per name, the number of possibilities is 203 (=8,000), a very large search space for just one name, now multiply it by the number of names on a watch list!
There are further challenges with the list method – how do you score matches when one of your 8,000 query variants matches more than one name in the database? It is also difficult to handle other types of variation, like nicknames, initials, and titles, without expanding the search space even more. A benefit of the list method is that it is simple to maintain. When a user complains about a missed match, it’s easily added to the name database.
However, easy maintenance may not be enough to offset the decreased speed. For applications with that require high-throughput over millions of names, such as watchlist screening, anti-money laundering (AML), and know your customer (KYC), this approach is likely to be too slow or require a lot of expensive hardware. Edit distance method Pros: Easy to implement Cons: Limited to Latin-based languages; all swaps are weighted evenly, missing linguistic nuances This approach looks at how many character changes it takes to get from one name to another.
“Cindy” and “Cyndi” have an edit distance of 1 since the “i” and “y” are merely transposed, whereas “Catherine” and “Katharine” have an edit distance of 2 as the “C” turns into a “K” and the first “e” becomes an “a.” Methods which look at the character-by-character distance between two names include the Levenshtein distance, the Jaro–Winkler distance, and the Jaccard similarity coefficient.
These approaches look at some combination of two factors (1) the number of similar characters and (2) the number of edit operations it takes to turn one name into the other—the operations being, insert, delete, and transpose. Although these comparisons are quick, they do not capture linguistic nuance.
All edits are given the same weight. Thus changing “c” to “p” is weighted equally as “c” to “k” although in English the latter substitution might more clearly indicate a similar name, as in “Catherine” vs. “Katherine.” Further, a one-to-many character mapping is not possible, as in the case of the Arabic character “sheen” ش which is frequently mapped to “sh” in English. And, just as with the common key method, a non-Latin script name must first be transliterated to Latin script before the comparison can be executed, as explained in the discussion of “The Weakness of the Common Key Method in Matching Across Scripts”.
Statistical similarity method Pros: Matches across languages and scripts; offers greater precision Cons: Slower performance; high barrier to entry as it requires training data and adjusting features etc.
A statistical approach takes hundreds, if not thousands, of matching name pairs and trains a model to recognize what two “similar names” look like so that the model can take two names and assign a similarity score. A statistical model that has been trained on thousands of pairs of matching names offers high accuracy and the ability to directly match names written in different languages without first transliterating names to Latin script.
This method has a higher barrier to entry, as collecting the matching names requires significant resources, but the accuracy may be well worth the effort.
A downside is the slowness of execution. A system only using the statistical method to sift through millions of names to look for matches may be too slow to be feasible in high-transaction environments. Word embedding method for organization names Pros: makes semantic matches that a spelling-centric method would miss Cons: only relevant to organization name matching Organization names differ from human names in that variations may include synonyms that look and sound entirely different than the target name.
In these cases, two names referring to one company are semantically similar but phonetically different. For example, a human can quickly infer that corporation, company, and group are all similar words often found in an organization’s name, but standard name matching techniques like the edit distance method would be unlikely to make the connection. In these cases, word embeddings can make the match.
Word embeddings are numerical vector representations of a word’s semantic meaning. If two words or documents have a similar embedding, they are semantically similar.
For example, the embeddings of “woman” and “girl” are close to one another in the vector space, meaning they are semantically similar. Contrastingly, the embeddings of “whale” and “philosophy” are far from one another because they are not semantically related.
Applied to organizations, the word embedding method recognizes that Eagle Drugs and Eagle Pharmaceuticals are most likely the same company. A two-pass, hybrid method: the best of breed Hybrid approaches backfill weakness in one approach with the strength of a different approach.
For example, hybrid approach may first use the common key method for high recall, and then put its results through the statistical method for greater precision. In the first pass— the faster common key method and high recall—winnows the candidate pool to a smaller set of likely matches.
This step is particularly vital when list has names in different languages, first transliterating them to —typically—English—before assigning metaphones. The second pass over the culled down list then uses a high-precision statistical method to filter the highest scoring matches to the top, making fine-grained distinctions between different matches.
Compared to the common key method alone, accuracy is greatly improved by this hybrid method. Instead of being locked into a coarse comparison of derived keys (for better or worse), the second pass of the hybrid approach takes a fresh look at the original names in their original scripts before scoring their similarity.
This hybrid method also avoids the weaknesses of the list approach by not relying on mass generation of name variations, but instead, uses (via the statistical model) the linguistic variations of names in each language. This linguistic knowledge of name variations also gives the hybrid approach an edge over the edit distance method, which cannot directly compare names in different scripts.
The result is a fast, accurate, .
Firstly, let’s consider what we mean by a company name, as CRM systems and other database structures often use the terminology of company name but in reality it’s a catch all phrase that relates to any organisational body that we are likely to do business with.
Sometimes other field descriptions are used such as Organisation Name, Entity Name, Business Name etc.. But does this really matter. Well if we are talking about matching and thinking in accordance with good data quality management then it does.
Legal Structures For instance a Company name is normally made of 2 parts, the unique recognisable name of the business and the type of legal entity that the business trades under. i.e. Microsoft Corporation / Tesco PLC / American Airlines Group Inc On occasion the company name will also provide additional entity type information such as Group, Bank, Holding Company etc..
Some organisation names will not include a legal identity element such as a Public Sector body, or a non-limited sole trader business. Does this matter? It’s quite typical that Matching solutions will either request you to provide a list of business entity types or will have this knowledge embedded with their technology and will look to strip this information out of the business name before any matching takes place. For example: ‘Bank of America Corporation’ would become Bank of America.
This type of data preparation is designed to help the matching engine make find the most appropriate matches by not focusing on generic keywords and terms that will appear many times within the data. However, removing the entity type is not always helpful. Sometimes, especially for larger organisations you will find that they have many subsidiaries which are of different legal structures and are registered as separate businesses on their own behalf.
Therefore the preferred match would be the part of the organisation with the same legal structure. Other types of business name entities. Using algorithms, artificial intelligence and logic specifically geared to the data in question will help to achieve the best possible results and simply the matching process.
For example if you are selling into the Hotel sector then using a data description of Hotel Name, rather than a generic company name would be more beneficial. With the logic / matching processes focused on identifying patterns, abbreviations and acronyms specifically focused to the Hotel sector.
The matching engine should be tuned to understand the nuances of the Hotel name, recognising that words and phrases like Hotel and B&B are going to be of less importance than recognised Hotel Chain names like Hilton and Best Western. The more specific the matching solution is for a particular type of entity name the more likely the better results you can achieve.
Abbreviations and Acronyms Another common issue to consider is abbreviations and acronyms. Its common place for business entities to use Acronyms alongside their company name, there are many examples of these, for example HP / Hewlett Packard, BT/British Telecom, GM/General Motors. Additionally we also have common terms that get abbreviated order to make data entry simpler, with ampersand’s (‘&’) used instead of ‘and’ and ltd to mean ‘limited’.
‘Corp’ for ‘Corporation’ etc. So we also need to have an extensive library of knowledge that can be called upon to help match identify and approve matches. However, it is not always straightforward as often there are exceptions o every rule.
For example the term Limited can be used to refer to a limited liability company, but it could also be part of the organisation name, for instance the business ‘Limited Brands Inc’.
So removing the term limited for this instance would result in poor matching. We also have other complexities to deal with as the same abbreviation can have multiple meanings, and this can vary by country. For instance BT would likely mean British Telecom when looking at UK data but in the Australia we have the company ‘BT Lawyers PTY LTD’, just one of many potential issues. So we need intelligent application of this extensive knowledge, one rule does not fit every scenario. Non Exact Matching Company Name Matching So the easy part of matching company names is generally exact matching, if the data is the same then it’s a match, but where it gets more complicated is when we have to make decisions on non-exact matches.
This is when we start to explore the use of fuzzy logic, computational algorithms, probabilistic reasoning and other artificial intelligent technologies and machine learning. However, with Fuzzy logic you solve one problem and create another. You see the fuzzy logic can generate likely match candidates and give them a score of likelihood, but eventually you have to make a decision, are they a match or not.
Process wise you either make that decision systematically on behalf of the users or you ask the user to make the decision themselves. If you ask the user to make the decision you typically add a large amount of work / time and hence cost to your matching project, if you make the decisions automatically then you have to choose a threshold or serious of thresholds to base your decision on and find a compromise you are happy with balancing the number of matches with the trustworthiness of the matches.
It is always the challenge of matching to balance quality with quantity, usually determined by cost effectiveness. With having users validating you’re more complex matching, and capturing there input you can provide essential input into machine learning processes, and help tip the balance of the quality / quantity seesaw in favour of the quality. Noise Words in the Company Name Another important consideration when matching company names is noise words.
It is not uncommon for company names to include many noise words that can be a distraction for the matching engine. For example European Headquarters of Ford Motor Company. Typical matching engines create a match key of a company name, often using the first 16 or so characters of the company name alongside some fuzzy logic processes to remove duplicate letters and using some phonetic based logic, to help speed up the process of finding good matches, but when noise words are included then this key approach is of limited use.
Our approach is very different, and our logical processes assess a company name and determine which elements are of most importance, helping us to find the most appropriate matches. Trading Names / Former Names and other considerations Many CRM systems and other databases will allow for business names to exist with additional insight, for instance in the D&B worldbase we see Former Name and upto 5 different trading names that a business trades under.
It is not uncommon for a business really to be only known to people via its trading name and people being unaware of the actual legal name the business is registered under. For example franchise business will often trade under a recognised brand name, but will be owned and operated by a different legal entity.
Former Names can also be very helpful when matching as your data may be more up to date than the data you are matching against and therefore by having the former name available can help match more data, more easily. Often matching systems will overlook this as it adds many more steps into the matching process and can be very time consuming without the necessary computing power to process this quickly.
Data Collaboration When you are looking to match a company name, then often you can make the process easier if you have additional information that can be used to help find the best matches. By using an address with your company name, you can help screen out lots of similar company names that are unlikely to be the same.
Website can also be helpful in this way as can telephone numbers, fax numbers, Long/Lat coordinates, postal codes, countries etc.. The more quality helpful information you can provide the matching engine generally the better results you will get. Summary Entity matching for company names has many considerations, we hope that this information will prove itself useful to you when choosing or building a solution for company name matching.
Match2Lists provides one of the most comprehensive and powerful matching solutions available today, using state of the art in memory parallel processing technology and an industry leading visual interface.
How to use Excel Index Match (the right way)